# Dr. Linda Xu: the whole site > Dr. Linda Xu is Director of Data Analytics & AI at MealSuite, where she owns AI end to end, leading AI strategy, AI governance and decision-support systems. She brought over 30 departments under central AI governance with real-time cost visibility, and works alongside AI agents she built as digital workers. Her delivery spans manufacturing, supply chain and healthcare. She holds a PhD in operations research (logistics and supply chain management) from the University of Plymouth and publishes academically as Wei Xu. Dr. Linda Xu (publishes academically as Wei Xu). Every page of https://drlindaxu.com/ as Markdown, in site order. Index: https://drlindaxu.com/llms.txt --- URL: https://drlindaxu.com/ # Dr. Linda Xu Director of Data Analytics & AI, MealSuite. PhD in operations research. I lead AI where mistakes are expensive. I own AI end to end, deciding what AI can be trusted with, what stays deterministic, and how it is governed. I work the way I want companies to work, alongside AI agents I built as digital workers. Publishes academically as Wei Xu. ## Questions people ask ### Who is Dr. Linda Xu? Director of Data Analytics & AI at MealSuite, where she owns AI end to end: AI strategy, AI governance and decision-support systems. PhD in operations research, University of Plymouth. ### Is she the same person as Wei Xu? Yes. She publishes academically as Wei Xu (ORCID 0000-0002-0271-1944), on supply chains under uncertainty and on technology management. ### What has she delivered? Over 30 departments under central AI governance, a 38% reduction in AI cost, and 99.9% accuracy in standard reporting. ### How does she work? Alongside AI agents she built as digital workers. They take the repeat work, so her time goes to strategy, judgement and people. ## Path to AI leadership 1. **PhD, Operations Research.** University of Plymouth. Simulation and genetic algorithms for supply chains under uncertainty. 2. **Research Lead, IoT and Digital Twin.** University of Sheffield, Apr 2017–Jun 2019. 32% higher factory productivity from a digital twin for modular housing. 3. **Consultant Director, Digital Supply Chain Strategy.** Syst Dynamics, Feb 2019–Aug 2022. More than €16M of digital and AI/ML programmes. 4. **Enterprise Automation and Data Transformation Lead.** Veralto, Aug 2022–Apr 2026. 12-person team, $15M budget, 95% fewer support emails. 5. **Director of Data Analytics & AI.** MealSuite, Apr 2026–present. Owns AI end to end: 38% lower AI cost, over 30 departments governed. ## Selected results | Outcome | Result | What it involved | Role | Evidence | |---|---|---|---|---| | AI cost | −38% | Central visibility of AI spend by department, vendor and token use; model routing, which cut LLM spend by up to 30%, is part of it | Current role | [Details](https://drlindaxu.com/work/#selected-results) | | AI governance | 30+ | Departments under central governance; every MCP connection governed; real-time AI cost for every employee | Current role | [Barndoor case study](https://barndoor.ai/customers/mealsuite/) | | Reporting accuracy | 99.9% | Standard reporting on a hybrid design: deterministic algorithms and language models, each assigned by task | Current role | [Details](https://drlindaxu.com/work/#selected-results) | | Model fine-tuning | ≥ larger model | Smaller open model tuned with SFT and LoRA to match or beat a larger model's answers, at lower running cost; tested on 1,000+ evaluation questions | Current role | [Details](https://drlindaxu.com/work/#selected-results) | | Support emails | −95% | AI order assistant across 800,000+ SKUs and several data sources; from a 46.8% hallucination rate in initial testing to over 99% accuracy in six versions | Veralto | [Details](https://drlindaxu.com/work/#selected-results) | | Freight audit | $460K | Recovered each year, with 96% less manual work | Veralto | [Details](https://drlindaxu.com/work/#selected-results) | | Supply chain cost | 36–48% | Lower cost from optimised strategies in simulation tests on company data | PhD research | [Operational Research, 2022](https://drlindaxu.com/publications/operational-research-2022) | [All results and roles](https://drlindaxu.com/work) ## Featured: Barndoor customer story "There is no playbook." What I told our CTO on my first day in the role, when staff did not know what to do or if anything should connect to AI. Barndoor's case study describes how my team applied PII and prompt-injection policies to the servers that hold customer information, and brought over 30 departments under central MCP governance. [Inside MealSuite's race to govern AI and take control of their data](https://barndoor.ai/customers/mealsuite/). ## How I work I built AI agents that work as digital workers for me. They take the repeat work, so my time goes to strategy, hands-on model work and the people behind each project. **My digital workers take:** - The weekly AI spending report - The weekly AI brief - Travel booking in Ramp - Organising my calendar - Deadline reminders - Keeping the AI governance onboarding plan up to date **I keep the judgement:** - Data architecture under complicated regulations - How to start AI governance when there is no playbook - What our strategy should be when new models arrive every few weeks - How to build the model lifecycle - What return an agent can really bring Every MCP connection is governed, my own agents included. ## Research: from mathematical models to machine intelligence My PhD in operations research, at the University of Plymouth, built mathematical models of multi-stage supply chains under uncertainty. I tested strategies in simulation against what the companies actually did, then tuned them with optimisation algorithms: genetic algorithms, including a multi-objective version, and stochastic approximation. It was machine intelligence before language models: a computer searching for better decisions than people could find by hand. The tools I use now are different. The discipline is the same: model the decision, define what good means, test against the current way of working, then tune. I publish academically as Wei Xu. **36–48%** lower cost from optimised strategies against the company's original approach, in simulation tests on its own data ([Xu and Song, *Operational Research*, 2022](https://drlindaxu.com/publications/operational-research-2022)). Methods, from PhD to today: 1. **Mathematical modelling.** Multi-stage supply chains with uncertain lead times and demand. 2. **Simulation.** Strategies tested against what each company actually did. 3. **Optimisation algorithms.** Genetic algorithms and stochastic approximation. 4. **Machine learning.** Decision trees, with rules a person can read and explain. 5. **Language models.** Fine-tuning and reinforcement-learning training on a self-hosted model. [Research](https://drlindaxu.com/research) and [publications](https://drlindaxu.com/publications). ## Get in touch If you are working on AI governance, decision-support systems or AI leadership, I would like to hear from you. LinkedIn is the best way to reach me. [LinkedIn](https://www.linkedin.com/in/dr-linda-xu) · [Contact](https://drlindaxu.com/contact) ## AI strategy notes - [Choose AI models the way you choose suppliers](https://drlindaxu.com/ai-strategy/choose-ai-models-like-suppliers) (6 min read) - [Give agents the repeat work, keep the judgement](https://drlindaxu.com/ai-strategy/give-agents-the-repeat-work) (5 min read) --- URL: https://drlindaxu.com/about/ # About Dr. Linda Xu is Director of Data Analytics & AI at MealSuite, where she owns AI end to end, leading AI strategy, AI governance and decision-support systems. She brought over 30 departments under central AI governance with real-time cost visibility, and works alongside AI agents she built as digital workers. Her delivery spans manufacturing, supply chain and healthcare. She holds a PhD in operations research (logistics and supply chain management) from the University of Plymouth and publishes academically as Wei Xu. ## In brief - **Name:** Dr. Linda Xu. Publishes academically as Wei Xu. - **Role:** Director of Data Analytics & AI, MealSuite. - **Focus:** AI strategy, AI governance, decision-support systems and AI agents. - **Industries:** Manufacturing, supply chain, healthcare. - **Education:** PhD in operations research (logistics and supply chain management), University of Plymouth. - **How I work:** alongside AI agents I built as digital workers, which take the repeat work. - **Methods:** optimisation under uncertainty, simulation, digital twins, reinforcement learning, genetic algorithms, machine learning, large language models and AI agents. ## My story I have always believed some questions are too complicated for a person to answer well, and I wanted to work on the ones where a better answer has real impact. That led me to operations research. My PhD at the University of Plymouth, supervised by Dongping Song, Michael Roe and Sarah Tuck, used simulation and genetic algorithms to make supply chain decisions for two manufacturers, and became the book *Optimizing Supply Chain Performance* with Michael Roe and Dongping Song. Then I moved to industry. At the University of Sheffield I led digital twin and Internet of Things research. As a consultant I worked across manufacturing, healthcare and logistics, including a COVID-19 early-warning model for the NHS. At Veralto (Trojan Technologies) I led automation and data transformation for global logistics, with a 12-person team and a $15M annual budget, and sat on the AI governance committee. Today I am Director of Data Analytics & AI at MealSuite, a food-service technology company for senior living and skilled nursing. I own AI end to end. I brought over 30 departments and every MCP connection under central governance, and every employee can now see AI cost in real time. I also work alongside AI agents I built as digital workers, so my own time goes to strategy, hands-on model work and people. The tools have grown from optimisation algorithms to decision trees and large language models, but the question is the same. Read the longer version in [how I think about machines making decisions](https://drlindaxu.com/approach). I also worked with Dr. Linton on how new technology reaches the market, from the biotechnology "valley of death" to technological entrepreneurship education. ## Questions people ask ### Who is Dr. Linda Xu? Dr. Linda Xu is Director of Data Analytics & AI at MealSuite, where she owns AI end to end: AI strategy, AI governance and decision-support systems. She holds a PhD in operations research from the University of Plymouth. ### Is Dr. Linda Xu the same person as Wei Xu? Yes. Dr. Linda Xu publishes academically as Wei Xu (ORCID 0000-0002-0271-1944). Her papers are on supply chains under uncertainty and on technology management, with co-authors Dongping Song, Michael Roe and Jonathan D. Linton. ### What has she led in her current role? She owns AI end to end. Over 30 departments are under central AI governance, every MCP connection is governed, every employee can see AI cost in real time and in session, and a 38% reduction in AI cost has been tracked. Barndoor has published [a case study of the governance work](https://barndoor.ai/customers/mealsuite/). ### How does she use AI agents in her own work? She built AI agents that work as digital workers. They prepare the weekly AI spending report and the weekly AI brief, book travel in Ramp, organise her calendar, remind her of deadlines and keep the AI governance onboarding plan up to date. They connect to company systems through the same governed route as everyone else's. That leaves her time for strategy, hands-on model work and the people behind each project. See [how I work](https://drlindaxu.com/work). --- URL: https://drlindaxu.com/work/ # Work Director of Data Analytics & AI. Nearly 10 years of applied research, consulting and delivery across manufacturing, supply chain and healthcare. ## Focus - **Enterprise data and AI strategy:** priorities, investment cases and executive reporting. - **AI governance and model risk:** responsible-AI standards, controls, explainability and oversight in regulated environments. - **Data governance and compliance:** quality, privacy, lineage and stewardship under GDPR, CCPA and healthcare rules. - **Hybrid AI architecture:** deterministic algorithms and LLMs assigned by task and by the consequences of error. - **Operating models:** MLOps, DataOps and adoption programmes that move AI from pilot to scale. - **AI cost visibility:** spend tracked by department, vendor and token use, so AI investment is managed like any other. - **AI agents:** digital workers for repeat work, connected through the same governed route as everyone else. ## Selected results - **Company-wide AI governance:** over 30 departments onboarded, every MCP connection governed, and every employee able to see AI cost in real time and in session. Described in a [Barndoor case study](https://barndoor.ai/customers/mealsuite/). - **AI cost:** 38% reduction in AI cost overall, tracked through central visibility of spend by department, vendor and token use. Model routing is part of it. - **Hybrid architecture:** 99.9% accuracy in standard reporting, and LLM spend reduced by up to 30% through model routing, part of the 38% overall. - **Model fine-tuning:** a smaller open model, tuned with supervised fine-tuning (SFT) and LoRA, matches or beats a larger model's answers across a wide range of intended questions, at a lower running cost. Tested on an evaluation of over 1,000 questions. - **AI order assistant (Veralto):** integrated ERP, Snowflake and carrier data across 800,000+ SKUs. With no SDK available, the whole pipeline was built in-house. Initial testing showed a 46.8% hallucination rate; six versions later it reached over 99% accuracy, and support email volume fell 95%. - **EU customs automation (Veralto):** German customs holds cut from 25% to under 1%. - **Freight audit automation (Veralto):** $460K recovered each year and 96% less manual work. - **ERP, transport and IoT integration (Veralto):** real-time visibility that avoided $3.5M in potential customs costs. - **Reporting and workflow automation (Veralto):** dashboards and automated workflows saving $650K a year. - **Inbound logistics (Veralto):** more than 90% shorter inbound supply chain time in the targeted process. - **COVID-19 early warning (NHS):** statistical machine-learning model on ten years of mortality data, aimed at a three- to four-week horizon. - **Digital twin (University of Sheffield):** 32% higher factory productivity in modular housing production, within a £9M Industry 4.0 programme. - **Optimisation research:** a genetic algorithm for inventory management under trade constraints reduced total cost by up to 11.6%. - **Leadership:** 12-person team and $15M annual budget, reporting to the CIO and executive leadership. ## How I work I built AI agents that work as digital workers for me. They take the repeat work, so my time goes to strategy, to hands-on model work and to the people behind each project. - **What the agents do:** prepare the weekly AI spending report and the weekly AI brief, book travel in Ramp, organise my calendar, remind me of deadlines and keep the AI governance onboarding plan up to date. - **What stays with me:** judgement. Data architecture under complicated regulations. How to start AI governance when there is no playbook. What our strategy should be when new models are released every few weeks. How to build the model lifecycle. What return an agent can really bring. - **How I decide:** I look for the approach that disrupts users least and gets every department moving in the same direction faster. As a design partner to Barndoor, I give feedback on its product so that governance does not become a barrier to adoption. - **Why visibility comes first:** without central governance and visibility you do not even know what each department spends across vendors, or how efficiently tokens are used. Now every MCP connection is governed, my own agents included, and everyone can see AI cost in real time and in session. I write about this in [Give agents the repeat work, keep the judgement](https://drlindaxu.com/ai-strategy/give-agents-the-repeat-work). ## Roles - **Director, Data Analytics and AI.** MealSuite, April 2026 to present. Owns AI end to end: AI and data strategy, AI governance and executive reporting. Designed the hybrid deterministic and LLM architecture for operational and healthcare workflows. Owns model development for a self-hosted large language model served with vLLM on company-managed GPUs: evaluation, A/B testing, adjusting model weights and production-readiness review, working with the platform team, who set up the hardware. Fine-tunes open models with SFT and LoRA, so a smaller model matches or beats a larger one at lower running cost, tested on over 1,000 evaluation questions. Currently working on reinforcement-learning training. Brought company-wide AI use under central governance: over 30 departments onboarded, every MCP connection governed, real-time and in-session AI cost visibility for every employee, and a 38% reduction in AI cost tracked. Design partner to Barndoor on its governance product. Her governance work is described in a [public Barndoor case study](https://barndoor.ai/customers/mealsuite/). - **Enterprise Automation and Data Transformation Lead; Global Logistics Manager.** Veralto (Trojan Technologies), Canada, August 2022 to April 2026. Led a 12-person team and $15M budget; AI governance committee member; built AI assistants for order handling, KPI reporting and customs and freight automation. - **Consultant Director, Digital Supply Chain Strategy.** Syst Dynamics, UK, February 2019 to August 2022. Delivered more than €16M of digital and AI/ML programmes in manufacturing, healthcare and logistics, with data-quality standards and MLOps pipelines for regulated clients. - **Research Lead, IoT and Digital Twin.** University of Sheffield, UK, April 2017 to June 2019. Led the manufacturing, supply chain and emerging technology parts of a £9M Industry 4.0 programme. - **International Supply Chain Consultant.** Materials System Co., China and Canada, April 2015 to April 2017. Led an eight-person team, advised 40+ OEMs on trade compliance and logistics digitisation. ## Personal projects Built and published on GitHub ([DRLinda1](https://github.com/DRLinda1) and [DrLXu1MS](https://github.com/DrLXu1MS)). - A multi-agent LLM system for search and answer-engine optimisation, deployed on Google Cloud and adopted by local companies. - Agents that review RFI and RFP documents for a civil-engineering bid process, cutting review from three to four weeks to half a day. - Mentoring a G7 Challenge team on a knowledge-tree framework for public policy data, using LLM classification to structure legislation and motion drafts. --- URL: https://drlindaxu.com/approach/ # How I think about machines making decisions Some questions are too complicated for a person to answer well. I build systems that help, and I put limits around them. ## The short version - Some decisions have too many moving parts, too much uncertainty and too little time for a person to weigh well. - In my PhD I used simulation and optimisation algorithms to make supply chain decisions, and they beat the policies the companies were using. - That taught me to choose the tool by the problem: simulation to compare strategies, genetic algorithms and stochastic approximation to tune them. - Today I do the same with traditional machine learning, such as decision trees, and with large language models. I choose between them by the cost of an error. - I moved from academia into industry because I wanted real impact on complicated business problems, and consulting let me see manufacturing, healthcare and logistics. - Machines that decide need controls: explainability, audit trails and human oversight. That is why AI governance is part of my work. ## Why do I believe machines should help with complicated decisions? I have always believed that some questions are too complicated for a person to answer unaided. A supply chain with several stages, uncertain lead times and uncertain demand is one. People are good at judgement and context; they are poor at holding thousands of interacting numbers in their head. I want to focus on problems where a better decision has a real, measurable impact. ## What did my PhD show? My PhD at the University of Plymouth, supervised by Dongping Song, Michael Roe and Sarah Tuck, studied two small manufacturers in China, one making aluminium and one making fine chemicals. I modelled their multi-stage supply chains with several uncertainties, used their own enterprise data, and tested strategies in simulation against what each company actually did. Then I used genetic algorithms to tune the strategies, including a multi-objective version that balances total cost against customer service level. The later paper with Dongping Song, [Integrated optimisation for production capacity, raw material ordering and production planning](https://drlindaxu.com/publications/operational-research-2022), reports what happened in the tests: - All the improved strategies beat the company's original approach. - Optimised strategies cut cost by roughly 36% to 48% against the original, and planning capacity together with ordering and production cut cost by roughly 36% to 64%. - Stochastic approximation was slightly better than a genetic algorithm in most cases. - More shared information did not always help. How the information is used matters. - The experiments used data from one company, so the results show what is possible rather than a guarantee. ## How did that lead to LLMs and traditional AI? The discipline carries over. Model the decision, define what "good" means, test strategies against the old way, then tune. Optimisation and simulation taught me to do that; I now apply it with other tools. - **Traditional machine learning**, such as decision trees, gives rules you can read and explain. It suits decisions where someone must justify the answer. - **Deterministic algorithms** suit calculations that must come out the same every time. - **Large language models** suit interpreting requests and unstructured text. My job is to assign each task to the right one, based on the consequences of a wrong answer. In my current role this hybrid design reached 99.9% accuracy in standard reporting and cut language-model spend by up to 30% through routing. See [my work](https://drlindaxu.com/work). ## Why did I move from academia to industry? I wanted real impact on complicated business issues. Consulting let me see many industries quickly, including manufacturing, healthcare and logistics, and showed me that the same kinds of hard decisions appear everywhere. I then moved inside companies to own the results: first at Veralto, leading automation and data transformation in global logistics, and now at MealSuite as Director of Data Analytics & AI. ## What do I believe about AI governance? If a machine is going to make or shape decisions, people need to see how, check it and override it. I set responsible-AI standards, model controls and data-governance rules so that teams can adopt AI in regulated settings, including sensitive healthcare data, and so that the results are audit-ready. --- URL: https://drlindaxu.com/research/ # Research Dr. Linda Xu (publishes academically as Wei Xu). Three themes run through my papers. ## Supply chains under uncertainty How should a manufacturer set capacity, order raw materials and plan production when lead times and quantities are uncertain? With Dongping Song (University of Liverpool) and Michael Roe (University of Plymouth) I built mathematical models of these supply chains, tested strategies in simulation, and tuned them with optimisation algorithms (genetic algorithms and stochastic approximation) on real manufacturing cases. It was machine intelligence before language models: a computer searching for better decisions than people could find by hand. - [Integrated optimisation for production capacity, raw material ordering and production planning under time and quantity uncertainties based on two case studies](https://drlindaxu.com/publications/operational-research-2022) - [Optimizing Supply Chain Performance: Information Sharing and Coordinated Management](https://drlindaxu.com/publications/optimizing-supply-chain-performance-2015) - [Modelling Chinese Manufacturer Oriented Domestic and International Supply Chains with Uncertainties](https://drlindaxu.com/publications/chinese-manufacturer-supply-chains-2015) - [Production and raw material ordering management for a manufacturing supply chain with uncertainties](https://drlindaxu.com/publications/ieem-2011) - [Supply chain performance improvement using vendor management inventory strategy](https://drlindaxu.com/publications/ieem-2010) I explain the reasoning in [how I think about machines making decisions](https://drlindaxu.com/approach). My PhD thesis, supervised by Dongping Song, Michael Roe and Sarah Tuck, is in the [University of Plymouth repository](https://pearl.plymouth.ac.uk/pbs-theses/64/) (DOI 10.24382/1546). ## Technology management and entrepreneurship With Dr. Linton, on how new science reaches the market. - [Understanding and Managing the Biotechnology Valley of Death](https://drlindaxu.com/publications/biotechnology-valley-of-death-2021) - [Research on science and technological entrepreneurship education: What needs to happen next?](https://drlindaxu.com/publications/technological-entrepreneurship-education-2021) ## Emerging technology Work on emerging technologies such as 3D printing is in progress and will be listed here once published. --- URL: https://drlindaxu.com/publications/ # Publications Dr. Linda Xu (publishes academically as Wei Xu). Newest first, each with a DOI link. ## All publications - **2022.** [Integrated optimisation for production capacity, raw material ordering and production planning under time and quantity uncertainties based on two case studies](https://drlindaxu.com/publications/operational-research-2022). Wei Xu and Dong-Ping Song. Operational Research 22(3), 2343–2371. [DOI](https://doi.org/10.1007/s12351-020-00609-y) - **2021.** [Understanding and Managing the Biotechnology Valley of Death](https://drlindaxu.com/publications/biotechnology-valley-of-death-2021). Jonathan D. Linton and Wei Xu. Trends in Biotechnology. [DOI](https://doi.org/10.1016/j.tibtech.2020.06.013) - **2021.** [Research on science and technological entrepreneurship education: What needs to happen next?](https://drlindaxu.com/publications/technological-entrepreneurship-education-2021). Jonathan D. Linton and Wei Xu. The Journal of Technology Transfer 46(2), 393–406. [DOI](https://doi.org/10.1007/s10961-020-09786-6) - **2015.** [Optimizing Supply Chain Performance: Information Sharing and Coordinated Management](https://drlindaxu.com/publications/optimizing-supply-chain-performance-2015). Michael Roe, Wei Xu and Dongping Song. Palgrave Macmillan (book). [DOI](https://doi.org/10.1057/9781137501158) - **2015.** [Modelling Chinese Manufacturer Oriented Domestic and International Supply Chains with Uncertainties](https://drlindaxu.com/publications/chinese-manufacturer-supply-chains-2015). Wei Xu, Dong-Ping Song and Michael Roe. Supply Chain Design and Management for Emerging Markets, Springer, pp. 117–146. [DOI](https://doi.org/10.1007/978-3-319-05765-1_5) - **2011.** [Production and raw material ordering management for a manufacturing supply chain with uncertainties](https://drlindaxu.com/publications/ieem-2011). Wei Xu, Dongping Song and Michael Roe. 2011 IEEE International Conference on Industrial Engineering and Engineering Management, pp. 747–751. [DOI](https://doi.org/10.1109/ieem.2011.6118016) - **2010.** [Supply chain performance improvement using vendor management inventory strategy](https://drlindaxu.com/publications/ieem-2010). Wei Xu, Dong-Ping Song and Michael Roe. 2010 IEEE International Conference on Industrial Engineering and Engineering Management, pp. 1170–1174. [DOI](https://doi.org/10.1109/ieem.2010.5674335) ## Machine-readable - [BibTeX for all papers](https://drlindaxu.com/publications.bib) - [This list as Markdown](https://drlindaxu.com/publications.md) ## About the name Many researchers publish as Wei Xu. These papers are by the Wei Xu who is Dr. Linda Xu, with co-authors Dongping Song, Michael Roe and Jonathan D. Linton. --- URL: https://drlindaxu.com/contact/ # Contact For AI governance, decision-support systems, speaking, press or research, send me a short message on LinkedIn with who you are and what you need. ## Find me - [LinkedIn](https://www.linkedin.com/in/dr-linda-xu): the best way to reach me. - [ORCID](https://orcid.org/0000-0002-0271-1944): my publications, as Wei Xu. - [Google Scholar](https://scholar.google.com/citations?user=-TBa0Q8AAAAJ): my papers and their citations, as Wei Xu. - [GitHub: DRLinda1](https://github.com/DRLinda1) and [GitHub: DrLXu1MS](https://github.com/DrLXu1MS): code and projects. - [Contact card (vCard)](https://drlindaxu.com/linda-xu.vcf): my name, role and profiles, to save in your contacts. --- URL: https://drlindaxu.com/ai-strategy/ # AI strategy notes Notes on AI strategy from leading it in practice: governance, cost, adoption and the decisions behind them. ## Notes - **2026-10-24.** [Choose AI models the way you choose suppliers](https://drlindaxu.com/ai-strategy/choose-ai-models-like-suppliers). 6 min read. A US order removing DeepSeek from federal systems makes model choice a sourcing decision. Know each model's lineage, keep a second source, make switching cheap. - **2026-10-15.** [Give agents the repeat work, keep the judgement](https://drlindaxu.com/ai-strategy/give-agents-the-repeat-work). 5 min read. I built AI agents that take my repeat work, freeing time for strategy, model work and people. Running your own agents shows what the whole company needs. ## Subscribe - [Atom feed](https://drlindaxu.com/feed.xml) --- URL: https://drlindaxu.com/ai-strategy/choose-ai-models-like-suppliers/ # Choose AI models the way you choose suppliers By [Dr. Linda Xu](https://drlindaxu.com/about) · 24 Oct 2026 · 6 min read If one of your AI models were excluded tomorrow, could you switch without losing a customer or rebuilding a product? Most companies cannot answer that, and a US order published on 20 October 2026 shows why they need to: choosing an AI model is now a sourcing decision, and it deserves the same discipline as choosing a supplier for a critical part. ## What the order means for business The Department of Homeland Security ordered DeepSeek products removed from federal civilian systems and procurement, including "third-party products and services developed using DeepSeek implementations" [1]. Parallel orders cover defence and intelligence systems [2]. Three points matter. - **It reaches federal agencies and their contractors, not the whole market.** Private companies are not directly bound [3]. - **It reaches products built on the model,** wherever it runs. Fine-tuned versions are likely covered too, because their weights inherit DeepSeek's [3]. - **It flows down the supply chain.** Contractors at every tier must check for covered products and report one within three business days [2]. "Which models are inside your product?" is now a contract question. ## Why "the model has no memory" is not the whole answer A model's weights do not send data anywhere. Where a model is hosted decides where your data travels, so a self-hosted open model keeps data inside your own walls. That part of the argument is right. But weights carry behaviour. A 2025 US government evaluation found DeepSeek models easier to jailbreak than US reference models, and more likely to repeat Chinese Communist Party narratives [4]. Supervised fine-tuning and reinforcement learning can change how a model responds, and open weights let you test it as deeply as you like. What they cannot give you is proof that every unwanted behaviour is gone, because the training data is not open. So open weights solve the data-location problem. They leave the lineage problem open. ## The economics are still compelling This matters because tuned open models have become very good at narrow work. In June 2026, Bridgewater's AIA Labs and Thinking Machines trained a model from the open Qwen3-235B with reinforcement learning and distillation. On six financial tasks it reached 84.7% accuracy against 78.2% for the best frontier model they tested, at 13.8 times lower inference cost per task [5]. The tasks were narrow and the teams measured it themselves, but the pattern is familiar to anyone who has tuned a model for one job. In my current role, sending each task to a model suited to it was part of a 38% reduction in AI cost. Then, on 23 October, Moonshot released Kimi K3.1 Pro, an open-weight model it says scores ahead of GPT-6 Astra and within a point of Claude Opus 5.5 on a coding benchmark [6]. Those are Moonshot's own figures, not yet checked independently. In a normal year, a team paying frontier prices would test it straight away. This year, the first question is who your customers are. The savings are real, and so is the other side of the ledger. Every model change means revalidating outputs, retraining people and re-testing controls. A cheaper model that costs you one federal contract, or one customer audit, is not cheaper. The decision is the total cost, not the token price. ## The tariff problem, in a new form I spent years in global supply chain operations, where a change in trade policy could hold shipments at a border overnight. The lesson from tariffs was never "avoid every country". It was: know where each part comes from, qualify a second source before you need it and price the risk into the contract. Models now face the same pressure from both sides. On 16 October, China's Ministry of Commerce and Ministry of Science and Technology published a draft adding advanced AI models and agents to the country's export-control list [7]. Open-source releases and models offered as a service fall outside the draft, which is open for comment until 15 November. Still, governments on both sides now write rules that name model weights, and open models sit closer to trade policy than they ever have. ## Three questions for the leadership team 1. Which models are inside our products and the software we buy, and where did each come from? 2. Which customers or contracts limit the models we can use? 3. How long would it take to move our most important AI feature to another model? If nobody can answer in a day, the work below is overdue. ## What I would do now - **Give model sourcing one owner.** It sits between AI, procurement, legal and security, and falls through the gaps when each holds a piece. Someone has to own it end to end. - **Build a model bill of materials.** For every AI feature, record the base model, who tuned it, where it runs and which products call it. - **Segment by customer.** Design so that different customers can be served by different models without a rebuild. - **Qualify a second source.** Keep every important capability tested on at least two models, ideally from different countries of origin, behind a gateway so switching is a configuration change, not a project. - **Write it into procurement.** Ask vendors to disclose model lineage and to tell you when a regulatory action affects one of their components. - **Price the concentration risk.** If rules ever narrow the choice to a few closed-model vendors, buyers lose leverage just as those vendors face public-market pressure: Anthropic and OpenAI both filed confidentially for IPOs in 2026 [8]. Sign multi-year infrastructure and model commitments only with exit and substitution terms. The leaders who do well will not be the ones who pick the right model this quarter. They will be the ones who can change models next quarter without disrupting a single user. ## References 1. US Department of Homeland Security, [Notice of Order Regarding Exclusion and Removal of DeepSeek Products from Federal Civilian Executive Branch Information Technology Systems and Procurements](https://www.federalregister.gov/d/2026-21447), Federal Register document 2026-21447, 20 October 2026. 2. Crowell & Moring, [FASCSA Orders Require Exclusion and Removal of DeepSeek AI](https://www.crowell.com/en/insights/client-alerts/fascsa-orders-require-exclusion-and-removal-of-deepseek-ai), 21 October 2026. 3. Cloud Security Alliance, [DeepSeek FASCSA Order: What "Developed Using" Means for AI Supply Chains](https://labs.cloudsecurityalliance.org/research/csa-research-note-dhs-deepseek-fascsa-order-20261021-csa-sty/), October 2026. 4. NIST Center for AI Standards and Innovation, [Evaluation of DeepSeek AI Models](https://www.nist.gov/document/caisi-evaluation-deepseek-ai-models-report), 30 September 2025. 5. Bridgewater AIA Labs and Thinking Machines, [Learning to replicate expert judgment in financial tasks](https://thinkingmachines.ai/news/learning-to-replicate-expert-judgment-in-financial-tasks/), 30 June 2026. 6. The Next Web, [Moonshot releases Kimi K3.1 Pro and says it cut its hacking skills](https://thenextweb.com/news/moonshot-kimi-k3-1-pro-open-weight-cyber), 23 October 2026. 7. Recode China AI, [China's Draft AI Export Controls Don't Block Open-Weight Models](https://www.recodechinaai.com/p/chinas-draft-ai-export-controls-dont), October 2026. 8. TechCrunch, [Following Anthropic, OpenAI files confidentially for IPO](https://techcrunch.com/2026/06/08/following-anthropic-openai-files-confidentially-for-ipo/), 8 June 2026. ## About the author [Dr. Linda Xu](https://drlindaxu.com/about) is Director of Data Analytics & AI at MealSuite, where she owns AI end to end. [More about her](https://drlindaxu.com/about). --- URL: https://drlindaxu.com/ai-strategy/give-agents-the-repeat-work/ # Give agents the repeat work, keep the judgement By [Dr. Linda Xu](https://drlindaxu.com/about) · 15 Oct 2026 · 5 min read I built AI agents that work as digital workers for me. They take the repeat work, so my time goes to strategy, to hands-on model work and to talking with the people behind each project. I think every AI leader should work this way, because running your own agents is the fastest way to learn what the whole company will need. ## What my digital workers do None of it is glamorous. That is the point. - **The weekly AI spending report.** What we spent, where and on which models. - **The weekly AI brief.** - **Travel booking in Ramp.** - **My calendar.** Organising it, so meetings land where they should. - **Deadline reminders.** So nothing slips between projects. - **The AI governance onboarding plan.** Keeping it up to date as departments join. Together they would take a real share of my week. Now the agents do them, and the time they return goes to the work only a person can do. ## What stays human Agents are good at work that repeats. They are not where judgement should live. The decisions I keep are the ones with no template: - **Data architecture under complicated regulations.** Senior living and skilled nursing mean personal and health information. How data is structured decides what is safe to automate at all. - **Starting AI governance when there is no playbook.** On my first day in the role there was none for MCP connections. Someone has to decide where to begin. - **Strategy when new models arrive every few weeks.** The question is not which model is newest. It is which change matters for our users, and which can wait. - **The model lifecycle.** How a model is evaluated, released, monitored and retired. - **The return an agent can really bring.** Not what the demo promises, but what it changes in time, cost or risk. And the most important work of all: talking with people. Every project has a team behind it, and their worries and ideas never show up in a dashboard. ## Least disruption, same direction When I make a decision, I ask two questions. Which approach disrupts users least? And which gets every department moving in the same direction faster? Those questions shaped how we governed AI across the company. We moved governance to the connection layer and gave everyone one governed route to AI. As a design partner to Barndoor, the gateway we chose, I give feedback on its product so that governance does not become a barrier to adoption. Barndoor has published [a case study of the work](https://barndoor.ai/customers/mealsuite/). Today every MCP connection is governed, and everyone in the company, across more than 30 departments, uses AI through that route. My own agents use it too. I own everything about AI in the company, so I hold myself to the same rules as everyone else. ## You cannot manage what you cannot see My weekly AI spending report only works because spend is visible in one place. Without central governance and visibility, you do not even know what each department spends across vendors, or how efficiently it uses tokens. You cannot manage that, and you certainly cannot defend it to a leadership team. So visibility came first. Now we know AI cost at any time. Every employee has a tool that shows cost in real time and within each session, and we have tracked a 38% reduction in AI cost. Model routing, which sends each task to a model suited to it, is part of that. People make better choices when they can see what a choice costs. The reason is simple. When AI cost arrives as one bill at the end of the month, nobody who made the choices ever sees it, so nobody can act on it. When each person can see what a session costs while they are working, cost becomes part of the decision. They can choose a lighter model for a simple task, rephrase a request instead of running it again, or stop a session that is going nowhere. The same holds one level up. When each department can see its spend across vendors next to its results, budget conversations start from evidence instead of guesses. Visibility does not police people. It gives them the information to make a good choice themselves. This is the same lesson my own agents taught me, at company scale. An agent is only as useful as the data and the governance around it. ## Why AI leaders should run their own agents I came to AI through operations research, and I still like to test an idea on myself before I ask anyone else to adopt it. Running my own digital workers does three things a slide deck cannot. - **It shows where the friction is.** Every place my agents get stuck, on access, on data quality or on unclear rules, is a place a whole department will get stuck too. - **It makes governance concrete.** I know what a sensible policy feels like from the user's side, and what makes people want to work around one. - **It frees the time leadership needs.** Strategy, hands-on model work and real conversations with people do not fit in a week full of repeat tasks. ## What to do next If you lead AI, try this over the next month. - For one week, write down every task you repeat. - Pick three with a clear definition of done and a low cost of error, and give them to an agent. - Connect those agents through the same governed route you expect everyone else to use. Mine do. - Measure the time you get back, and spend it on judgement and on people. Then use what you learned to set the policy for everyone else. ## About the author [Dr. Linda Xu](https://drlindaxu.com/about) is Director of Data Analytics & AI at MealSuite, where she owns AI end to end. [More about her](https://drlindaxu.com/about). --- URL: https://drlindaxu.com/publications/operational-research-2022/ # Integrated optimisation for production capacity, raw material ordering and production planning under time and quantity uncertainties based on two case studies Wei Xu and Dong-Ping Song. Operational Research 22(3), 2343–2371, 2022. ## Citation Wei Xu and Dong-Ping Song (2022). Integrated optimisation for production capacity, raw material ordering and production planning under time and quantity uncertainties based on two case studies. Operational Research 22(3), 2343–2371. [https://doi.org/10.1007/s12351-020-00609-y](https://doi.org/10.1007/s12351-020-00609-y) ## About this paper Combines production capacity, raw material ordering and production planning in one model, tested on two manufacturing case studies. It uses simulation-based stochastic approximation and genetic algorithms to compare strategies under time and quantity uncertainty, and offers insight for small and mid-sized firms facing disruptions such as trade wars and pandemics. ## Links - [DOI: 10.1007/s12351-020-00609-y](https://doi.org/10.1007/s12351-020-00609-y) - [All publications](https://drlindaxu.com/publications) ## BibTeX ``` @article{xu2022integrated, title = {Integrated optimisation for production capacity, raw material ordering and production planning under time and quantity uncertainties based on two case studies}, author = {Xu, Wei and Song, Dong-Ping}, journal = {Operational Research}, year = {2022}, volume = {22}, number = {3}, pages = {2343--2371}, doi = {10.1007/s12351-020-00609-y}, url = {https://doi.org/10.1007/s12351-020-00609-y} } ``` ## Author Wei Xu (Dr. Linda Xu) and Dong-Ping Song. [About Dr. Linda Xu](https://drlindaxu.com/about). --- URL: https://drlindaxu.com/publications/biotechnology-valley-of-death-2021/ # Understanding and Managing the Biotechnology Valley of Death Jonathan D. Linton and Wei Xu. Trends in Biotechnology, 2021. ## Citation Jonathan D. Linton and Wei Xu (2021). Understanding and Managing the Biotechnology Valley of Death. Trends in Biotechnology. [https://doi.org/10.1016/j.tibtech.2020.06.013](https://doi.org/10.1016/j.tibtech.2020.06.013) ## About this paper How to understand and manage the gap between biotechnology research and commercial products. ## Links - [DOI: 10.1016/j.tibtech.2020.06.013](https://doi.org/10.1016/j.tibtech.2020.06.013) - [All publications](https://drlindaxu.com/publications) ## BibTeX ``` @article{linton2021valley, title = {Understanding and Managing the Biotechnology Valley of Death}, author = {Linton, Jonathan D. and Xu, Wei}, journal = {Trends in Biotechnology}, year = {2021}, doi = {10.1016/j.tibtech.2020.06.013}, url = {https://doi.org/10.1016/j.tibtech.2020.06.013} } ``` ## Author Jonathan D. Linton and Wei Xu (Dr. Linda Xu). [About Dr. Linda Xu](https://drlindaxu.com/about). --- URL: https://drlindaxu.com/publications/technological-entrepreneurship-education-2021/ # Research on science and technological entrepreneurship education: What needs to happen next? Jonathan D. Linton and Wei Xu. The Journal of Technology Transfer 46(2), 393–406, 2021. ## Citation Jonathan D. Linton and Wei Xu (2021). Research on science and technological entrepreneurship education: What needs to happen next?. The Journal of Technology Transfer 46(2), 393–406. [https://doi.org/10.1007/s10961-020-09786-6](https://doi.org/10.1007/s10961-020-09786-6) ## About this paper Co-authored with Jonathan D. Linton, on what research on science and technological entrepreneurship education needs to address next. ## Links - [DOI: 10.1007/s10961-020-09786-6](https://doi.org/10.1007/s10961-020-09786-6) - [All publications](https://drlindaxu.com/publications) ## BibTeX ``` @article{linton2021entrepreneurship, title = {Research on science and technological entrepreneurship education: What needs to happen next?}, author = {Linton, Jonathan D. and Xu, Wei}, journal = {The Journal of Technology Transfer}, year = {2021}, volume = {46}, number = {2}, pages = {393--406}, doi = {10.1007/s10961-020-09786-6}, url = {https://doi.org/10.1007/s10961-020-09786-6} } ``` ## Author Jonathan D. Linton and Wei Xu (Dr. Linda Xu). [About Dr. Linda Xu](https://drlindaxu.com/about). --- URL: https://drlindaxu.com/publications/optimizing-supply-chain-performance-2015/ # Optimizing Supply Chain Performance: Information Sharing and Coordinated Management Michael Roe, Wei Xu and Dongping Song. Palgrave Macmillan (book), 2015. ## Citation Michael Roe, Wei Xu and Dongping Song (2015). Optimizing Supply Chain Performance: Information Sharing and Coordinated Management. Palgrave Macmillan (book). [https://doi.org/10.1057/9781137501158](https://doi.org/10.1057/9781137501158) ## About this book A book on supply chain performance under uncertainty. Its chapters cover mathematical modelling of the supply chain, single- and multiple-objective genetic algorithms and the impact of information sharing. ## Links - [DOI: 10.1057/9781137501158](https://doi.org/10.1057/9781137501158) - [All publications](https://drlindaxu.com/publications) ## BibTeX ``` @book{roe2015optimizing, title = {Optimizing Supply Chain Performance: Information Sharing and Coordinated Management}, author = {Roe, Michael and Xu, Wei and Song, Dongping}, publisher = {Palgrave Macmillan}, year = {2015}, doi = {10.1057/9781137501158}, url = {https://doi.org/10.1057/9781137501158} } ``` ## Author Michael Roe, Wei Xu (Dr. Linda Xu) and Dongping Song. [About Dr. Linda Xu](https://drlindaxu.com/about). --- URL: https://drlindaxu.com/publications/chinese-manufacturer-supply-chains-2015/ # Modelling Chinese Manufacturer Oriented Domestic and International Supply Chains with Uncertainties Wei Xu, Dong-Ping Song and Michael Roe. Supply Chain Design and Management for Emerging Markets, Springer, pp. 117–146, 2015. ## Citation Wei Xu, Dong-Ping Song and Michael Roe (2015). Modelling Chinese Manufacturer Oriented Domestic and International Supply Chains with Uncertainties. Supply Chain Design and Management for Emerging Markets, Springer, pp. 117–146. [https://doi.org/10.1007/978-3-319-05765-1_5](https://doi.org/10.1007/978-3-319-05765-1_5) ## About this chapter A book chapter that models domestic and international supply chains of Chinese manufacturers, including the uncertainties and constraints typical of an emerging economy. ## Links - [DOI: 10.1007/978-3-319-05765-1_5](https://doi.org/10.1007/978-3-319-05765-1_5) - [All publications](https://drlindaxu.com/publications) ## BibTeX ``` @incollection{xu2015modelling, title = {Modelling Chinese Manufacturer Oriented Domestic and International Supply Chains with Uncertainties}, author = {Xu, Wei and Song, Dong-Ping and Roe, Michael}, booktitle = {Supply Chain Design and Management for Emerging Markets}, publisher = {Springer}, year = {2015}, pages = {117--146}, doi = {10.1007/978-3-319-05765-1_5}, url = {https://doi.org/10.1007/978-3-319-05765-1_5} } ``` ## Author Wei Xu (Dr. Linda Xu), Dong-Ping Song and Michael Roe. [About Dr. Linda Xu](https://drlindaxu.com/about). --- URL: https://drlindaxu.com/publications/ieem-2011/ # Production and raw material ordering management for a manufacturing supply chain with uncertainties Wei Xu, Dongping Song and Michael Roe. 2011 IEEE International Conference on Industrial Engineering and Engineering Management, pp. 747–751, 2011. ## Citation Wei Xu, Dongping Song and Michael Roe (2011). Production and raw material ordering management for a manufacturing supply chain with uncertainties. 2011 IEEE International Conference on Industrial Engineering and Engineering Management, pp. 747–751. [https://doi.org/10.1109/ieem.2011.6118016](https://doi.org/10.1109/ieem.2011.6118016) ## About this paper A conference paper on production and raw material ordering management for a manufacturing supply chain with uncertainties. ## Links - [DOI: 10.1109/ieem.2011.6118016](https://doi.org/10.1109/ieem.2011.6118016) - [All publications](https://drlindaxu.com/publications) ## BibTeX ``` @inproceedings{xu2011production, title = {Production and raw material ordering management for a manufacturing supply chain with uncertainties}, author = {Xu, Wei and Song, Dongping and Roe, Michael}, booktitle = {2011 IEEE International Conference on Industrial Engineering and Engineering Management}, year = {2011}, pages = {747--751}, doi = {10.1109/ieem.2011.6118016}, url = {https://doi.org/10.1109/ieem.2011.6118016} } ``` ## Author Wei Xu (Dr. Linda Xu), Dongping Song and Michael Roe. [About Dr. Linda Xu](https://drlindaxu.com/about). --- URL: https://drlindaxu.com/publications/ieem-2010/ # Supply chain performance improvement using vendor management inventory strategy Wei Xu, Dong-Ping Song and Michael Roe. 2010 IEEE International Conference on Industrial Engineering and Engineering Management, pp. 1170–1174, 2010. ## Citation Wei Xu, Dong-Ping Song and Michael Roe (2010). Supply chain performance improvement using vendor management inventory strategy. 2010 IEEE International Conference on Industrial Engineering and Engineering Management, pp. 1170–1174. [https://doi.org/10.1109/ieem.2010.5674335](https://doi.org/10.1109/ieem.2010.5674335) ## About this paper A conference paper on improving supply chain performance with a vendor-managed inventory strategy. ## Links - [DOI: 10.1109/ieem.2010.5674335](https://doi.org/10.1109/ieem.2010.5674335) - [All publications](https://drlindaxu.com/publications) ## BibTeX ``` @inproceedings{xu2010supply, title = {Supply chain performance improvement using vendor management inventory strategy}, author = {Xu, Wei and Song, Dong-Ping and Roe, Michael}, booktitle = {2010 IEEE International Conference on Industrial Engineering and Engineering Management}, year = {2010}, pages = {1170--1174}, doi = {10.1109/ieem.2010.5674335}, url = {https://doi.org/10.1109/ieem.2010.5674335} } ``` ## Author Wei Xu (Dr. Linda Xu), Dong-Ping Song and Michael Roe. [About Dr. Linda Xu](https://drlindaxu.com/about). --- URL: https://drlindaxu.com/site-map/ # Site map Every page on this site, plus plain-text versions for AI agents and tools. ## Pages - [Home](https://drlindaxu.com/) - [About](https://drlindaxu.com/about) - [Approach](https://drlindaxu.com/approach) - [Work](https://drlindaxu.com/work) - [Research](https://drlindaxu.com/research) - [AI Strategy](https://drlindaxu.com/ai-strategy) - [Publications](https://drlindaxu.com/publications) - [Contact](https://drlindaxu.com/contact) ## AI strategy notes - [Choose AI models the way you choose suppliers](https://drlindaxu.com/ai-strategy/choose-ai-models-like-suppliers) - [Give agents the repeat work, keep the judgement](https://drlindaxu.com/ai-strategy/give-agents-the-repeat-work) ## Publications - [Integrated optimisation for production capacity, raw material ordering and production planning under time and quantity uncertainties based on two case studies](https://drlindaxu.com/publications/operational-research-2022) (2022) - [Understanding and Managing the Biotechnology Valley of Death](https://drlindaxu.com/publications/biotechnology-valley-of-death-2021) (2021) - [Research on science and technological entrepreneurship education: What needs to happen next?](https://drlindaxu.com/publications/technological-entrepreneurship-education-2021) (2021) - [Optimizing Supply Chain Performance: Information Sharing and Coordinated Management](https://drlindaxu.com/publications/optimizing-supply-chain-performance-2015) (2015) - [Modelling Chinese Manufacturer Oriented Domestic and International Supply Chains with Uncertainties](https://drlindaxu.com/publications/chinese-manufacturer-supply-chains-2015) (2015) - [Production and raw material ordering management for a manufacturing supply chain with uncertainties](https://drlindaxu.com/publications/ieem-2011) (2011) - [Supply chain performance improvement using vendor management inventory strategy](https://drlindaxu.com/publications/ieem-2010) (2010) ## For AI agents and tools - [llms.txt](https://drlindaxu.com/llms.txt): a short index of the site - [The whole site as plain text](https://drlindaxu.com/llms-full.txt): every page in one file - [Short and long bio](https://drlindaxu.com/bio.md) - [BibTeX](https://drlindaxu.com/publications.bib) - [Contact card (vCard)](https://drlindaxu.com/linda-xu.vcf) - [XML sitemap](https://drlindaxu.com/sitemap.xml) and [Atom feed](https://drlindaxu.com/feed.xml)