Lisa Thomas
State Farm
Bloomington,IL
21
years of professional experience
A synthesis of the views I have been building across my emails, briefs, and project work. Two shifts run through all of it: the workforce function has to move from reactive to predictive, and our research function has to move from analyst to researcher. Both are responses to the same force — AI is collapsing the price of description, so the value moves to judgment.
Work History
Senior Statistician
21 Years 3 Months- Developed statistical models supporting claims forecasting, pricing analysis, and risk assessment.
- Analyzed large insurance datasets using SAS, and excel to improve decision accuracy.
- Validated model assumptions, documentation, and outputs for regulatory and business review.
- Collaborated with actuarial, underwriting, and product teams to translate findings into recommendations.
- Improved data quality through rigorous sampling, cleansing, and anomaly detection procedures.
Education
Ph.D. - Industrial And Organizational Psychology
Master of Science - Industrial And Organizational Psychology
Skills
Machine learning techniques
Data visualization expertise
Data collection methodologies
Data trends
Data audits
Multivariate analysis
Experimental design
Data mining techniques
Big data analytics
Natural language processing
Predictive modeling
SAS
Decision trees
Factor analysis
Cluster analysis
Problem-solving
Personal Information
Title: Enterprise Research / Organizational Intelligence
Section name
- Where the Work Is Going
- Part One — The Future of Work
- The model isn’t broken. The approach is.
- The way most organizations manage workforce capacity today is broken — not for lack of effort, but because of a lag in approach. We measure what happened instead of anticipating what will happen. We run annual planning cycles in a world that now changes quarterly. And we sit on mountains of workforce data without the analytical infrastructure to turn it into decision-ready intelligence. The fix is not more effort inside the old model. It is a different model.
- Reactive HR is a liability. Predictive foresight is the move.
- My central argument is a shift from reactive HR to predictive workforce foresight, with PTO and absence analytics as the entry point. When we integrate leave data with engagement signals, scheduling patterns, and workforce demographics, we stop managing absences after the fact and start leading the business with confidence. Organizations that make this shift see measurable gains in coverage efficiency, retention, and financial performance. Those that do not keep managing workforce risk one costly surprise at a time. The opportunity is clear; the only real question is whether we move now or wait until the cost of inaction is unavoidable.
- Human + Digital is an operating model, not a slogan.
- The future is not humans or machines — it is humans plus digital, with the roles drawn deliberately. The model is the digital guardrail: it sets the defensible ceiling, surfaces the pattern, flags the risk. The supervisor is the human judgment layer: context, exceptions, the call the data can’t make. Automation handles execution; people handle judgment, creativity, and relationships. The decisive variable is never the tool itself — it is the intentional human design of the system the tool sits inside.
- The real risk is the readiness gap.
- Across every credible source I track — Microsoft’s Future of Work work, SHRM, Gartner, and the AI labs’ own research — the same pattern holds: AI deployment speed is outpacing organizational, psychological, and governance readiness. The danger isn’t the technology. It’s adopting faster than we can govern, and accumulating risk we can’t see. A few realities I keep returning to:
- Investment is outrunning value. The spend on AI is real; the demonstrated organizational return lags behind the promise. Diffusion takes longer than procurement.
- Engagement is at a historic low. Gallup’s 2026 State of the Global Workplace report puts global engagement at 20% in 2025 — down from a 23% peak in 2022, the first back-to-back annual decline on record — with low engagement costing the world economy an estimated $10 trillion, about 9% of GDP. That is the backdrop to all of this; tools layered onto a disengaged workforce don’t fix the workforce.
- Displacement looks like contraction, not collapse. The near-term signal is hiring slowdowns and a thinning of cognitive-routine work, not mass unemployment — which is exactly why it’s easy to underreact to.
- Skills are stratifying. There is a real and growing wage premium for people who can direct, validate, and stress-test AI. The gap between AI-fluent and AI-adjacent widens every quarter.
- Procedural justice is not optional.
- Any change that touches people’s time, schedules, or secured commitments lives or dies on organizational justice — procedural, informational, and interactional. You can have the cleanest analytics in the building and still fail if the process feels arbitrary, the rationale is hidden, or the rollout treats people as inputs. Reclaiming secured time without a justice frame generates resistance no model can outrun. Fairness is a design requirement, not a communications afterthought.
- Build the thing. Don’t just recommend it.
- My bias is toward tangible proof — working models, one-page briefs, decks leadership can act on — over conceptual recommendations. A forecast you can open and pressure-test changes a conversation in a way a slide titled “we should be more predictive” never will. The future of work belongs to the people who can show it running, not just describe it.
- Part Two — The Future of Research
- If we stay analysts, we get automated.
- Here is the blunt version: if we don’t evolve from data analysts into researchers, AI will replace us — and leadership will notice the cost difference before they notice the quality gap. Pulling, cleaning, and visualizing data, building dashboards, summarizing survey results — that work is already being absorbed, fast. If our value proposition is “we turn data into reports,” we are competing with a tool that costs pennies and never sleeps. The “analyst” framing reinforces a commodity skill set at the exact moment the commodity is collapsing in price.
- Research is the part that doesn’t commoditize.
- Researchers do something a model can’t. Researchers decide what question is worth asking. They design the study before touching the data. They know the difference between a finding and a fluke, between correlation and causation, between what the numbers say and what the numbers mean. They defend their methods and tie evidence to decisions. AI can run the analysis — it cannot tell you whether the construct you measured is the construct that matters, sit across from an executive and explain why an engagement dip in property complex is a leading indicator rather than a morale problem, or judge when a sample is too small and a recommendation too risky to make.
- We already have the foundation. What we need is the posture.
- This is not about new titles or more headcount. We have the IO psychology training, the methodological rigor, the construct discipline — it has simply been undersold internally. The gap is posture. We too often accept requests like analysts: fast turnaround, descriptive answer, whatever was asked for. Researchers push back and reframe the question. Closing that gap is a change in how we show up, not a reorganization.
- Make the artifacts look like research.
- The practical move is changing what shows up in front of leadership. Study charters before work begins. Hypotheses stated upfront. Methods sections that name their limitations. Findings tied to the decision they inform. A visible annual research agenda instead of a queue of ad hoc requests. When the artifacts look like research, the team gets treated like researchers — and the conversation shifts from “can a model do this” to “who designs and governs what the models are asked to do.”
- AI fluency is part of the craft now.
- Research thinking and AI fluency are not a trade-off; the resilient profile combines both. A researcher who can’t direct, validate, and stress-test what a model produces is also exposed — just on a slower timeline. I think about team capability across four domains: research methods, executive translation, domain fluency (insurance operations, WFM, subrogation, property complex), and AI and data literacy. The goal is the person who runs three studies a quarter with AI as their research assistant — not the person the assistant replaces.
- Rigor is the moat.
- In an environment where anyone can generate a confident-sounding answer in seconds, verifiable rigor is the differentiator. The standards I hold are deliberate: inference is always distinguished from evidence, every cited source resolves to a live page, corrections are made visible rather than quietly applied, and claims are tiered by source quality. That discipline is also why I treat synthetic data as both a technical and a trust asset — done right, it lets us study sensitive workforce questions without exposing real people, and the trust comes from the transparency of the method.
- My own edge.
- Concretely, the value I add is interpretation: turning data into insight, looking at the same information differently from the room, challenging the assumption everyone has accepted, and asking the follow-up question others miss. Those are not analyst tasks. They are the reason the research framing protects the work.
- The Through-Line
- Both shifts answer the same pressure. AI is making description cheap — cheap forecasts, cheap dashboards, cheap summaries. So value migrates to the things description can’t cover: anticipating what’s coming, designing the system humans and machines share, deciding what is worth knowing, and standing behind the answer. The workforce function moves from reactive to predictive. The research function moves from analyst to researcher. Same logic, two fronts.
- The work that survives is the work that decides what is worth knowing. That has always been research. It is simply more obvious now.
- What this is drawn from
- This synthesis pulls together positions I have already stated across the following:
- Email / brief: “The case for switching from reactive to predictive analytics to examine PTO” and “From Reactive HR to Predictive Foresight — PTO Analytics and Workforce Planning” (June 2026).
- Work: Human + Digital staffing model and leadership brief for Subrogation and Property Complex (Verint–Workday seam; model as guardrail, supervisor as judgment).
- Work: Future-of-work research synthesis drawing on Gartner, Gallup, Deloitte, WEF, McKinsey, Microsoft, and SHRM; and the cross-report pattern analysis (Microsoft New Future of Work 2025, SHRM AI 2026, AI-lab interpretability research).
- Work: “Evolving from data analysts to researchers” argument and the IO Research Team capability map (research methods, executive translation, domain fluency, AI and data literacy).
- Standing standards: evidence-versus-inference discipline, verifiable sourcing, visible corrections, and the synthetic-data research agenda.
- Prepared with AI assistance and reviewed against my own correspondence and project files. Figures cited from external reports should be validated against the primary source before external use.
Timeline
Senior Statistician
State Farm
06.2005 - 09.2026Read More
University of Illinois At Urbana-Champaign
Ph.D. from Industrial And Organizational Psychology
Read More
University of Illinois At Urbana-Champaign
Master of Science from Industrial And Organizational Psychology
Read More