Data-AI Strategy
Data strategy and AI strategy are the same work. Written separately, the AI plan assumes data the organization does not have, and the data plan never connects to what leadership actually wants.
We work with your executives and the people closest to the data to produce four things: a shared statement of the problem, a North Star use case the organization can agree on, a scored assessment of where you stand today, and a prioritized set of recommendations with owners and timeframes. The work is done remotely, with your team involved throughout.
Why data and AI strategy belong together
AI has not replaced the data agenda. It has raised the stakes, changed which data counts as strategic, and turned governance into something that determines whether projects ship.
AI runs on context
A model is a general capability. What makes it useful to you is the context you can put in front of it: your protocols, your registries, your notes, your definitions. The ceiling on your AI is the reach and quality of the data you can safely deliver to it. That ceiling is set by data strategy, whether or not anyone wrote one.
AI changed what counts as strategic data
For decades we curated structured data for reporting. Most institutional knowledge never made it in. It lives in documents, narratives, correspondence, and images. Modern models can work with that material at scale, so a data strategy scoped only to the warehouse now covers a small fraction of what you have.
Every AI decision is a data decision
Can this leave our tenant? Whose consent covers it? Is PHI in scope? Which version is the system of record? These questions stall AI pilots late, after the model already works. Every one of them is a data governance question that should have been settled first.
So we write one strategy, in which the AI ambitions justify the data investments and the data investments make the AI ambitions achievable.
Alignment is the whole game
The organizations that struggle most with data are rarely the ones with weak people or small ambitions. Usually it is the opposite. The talent is real, the vision is wide, and the work still does not add up. Effort behaves like a wave, and waves that arrive out of phase cancel.
Destructive interference
- Excellent teams, each doing real work on a roadmap that makes sense on its own terms.
- A wide executive vision that no one disagrees with and no one has turned into a budget.
- Frustrated customers who cannot tell you what they want, because no one has shown them what they should be able to ask for.
- Effort moving in many directions at once, which cancels out at the organizational level.
Constructive interference
- One strategic priority that connects executive vision to where the budget and headcount actually go.
- Sequenced investment, so foundational work is funded because everyone can see what it unblocks.
- Customers who know what is coming and when, and who have learned which questions are worth asking.
- Work that compounds, because each team's output is the next team's input.
Why this is so hard to see from the inside
Destructive interference does not look like failure. Every team can show you good work, every leader can point to a win, and every roadmap defends itself on its own terms. The loss is hard to see because it sits between the efforts rather than inside any one of them. It shows up only in aggregate: initiatives that never quite compound, the same data rebuilt three times, and stakeholders who have quietly stopped asking.
Nothing in that picture is fixed by better people or a bolder vision. You already have both. What is missing is phase.
Same teams. Same budget. Same vision. Different phase. Getting an organization into phase is what a data-AI strategy is for, and that requires something concrete for people to align to.
The North Star Use Case
Most data roadmaps stall because every department has its own list and nothing is anyone's priority. The North Star addresses that. It is a single use case, broad enough that no one team can deliver it alone, stated in one sentence that a board member and a data engineer can both repeat. Below is the oncology version. We write one for your organization, in your language.
One question that no single department can answer
A good North Star spans functions deliberately, so that collaboration is the only way to reach it. It should hold the organization's attention for at least two years, and it should break into components that each deliver value on their own. In oncology it usually takes the form of a question the organization cannot currently answer:
Now look at what happened. Every highlighted term in that sentence is a core data element. We didn't start with a list of data projects and try to decide which ones mattered most. The North Star told us what data mattered and why. A long list of competing data priorities became seven concrete things to get right, each tied to a source and a clear purpose:
Disease
- Site, histology
- Tumor markers
Stage
- Pathologic stage
- Clinical stage
Comorbidities
- External medical records
- Patient self-report
- Problem list
Molecular
- Outside medical records
- Discretized results from external labs
Treatments
- Drug classification — immuno vs. chemo vs. hormone
- Lines of therapy, regimens
Response
- RECIST
- Radiomics analysis
Outcomes
- Survival, cause of death
- Progression, recurrence
- Quality of life
Once it exists, the North Star does real work:
- It gives teams across the organization a single shared purpose.
- It identifies the 5–10 core data elements that matter most to achieving the North Star.
- It becomes the test for proposed investments: does this move the North Star?
- It shows where the answer is blocked. In this example, most of the difficult elements arrive as narrative text and outside records, which is where AI can do real work.
An adoption model, scored together
Before recommending anything, we place your organization on an analytics adoption model and score it. You rate yourselves, we rate you, and we reconcile the differences together. That conversation is often one of the most useful in the engagement. Below is the Community Impact Analytics Adoption Model, our adaptation of the public-domain Healthcare Analytics Adoption Model for mission-driven organizations. We adapt the model to your sector.
The pattern we find most often: top-heavy
The model is progressive. Each level builds on the ones beneath it, but organizations can and do execute at the upper levels while under-investing below. Talented analysts produce genuinely good work at Level 4, and leadership concludes the data program is healthy. Often it isn't. Each of those wins is hand-built, hard to repeat, and more expensive than it looks. Naming that pattern, with evidence, is usually what unlocks funding for the less visible work at Levels 2 and 3.
AI makes this more visible. Pilots work well on curated extracts and then stall on the way to production, because production requires the governance, lineage, and access control that were skipped.
Four phases
The work is remote and developed together with your executive sponsor and stakeholders. We set the schedule with you at the outset, around your calendar and the availability of the people we need to talk to. Phases can overlap.
Initiate
- Identify the executive sponsor, point of contact, and stakeholder set
- Set the meeting cadence
- Draft and ratify the Chief Concern and the strategy vision
- Finalize the plan and schedule interviews
Inspect & Interpret
- Review org structures, roadmaps, and the current data environment
- 30–45 minute stakeholder interviews across levels and functions
- Score the adoption model together
- Deliver spot training on AI tools as real workflows surface
- Mid-engagement checkpoint and interim report
Integrate
- Synthesize themes into gaps tied to the North Star
- Draft recommendations and review with the sponsor
- Iterate, prioritize, and finalize
Inform
- Finalize the written report and executive presentation
- Present, or support your presentation of, the findings
- Run a short stakeholder survey on the strategy and priorities
- Deliver an addendum capturing the feedback
What the report contains
The engagement produces a written report and an executive presentation. You receive a full and irrevocable license to both. These are the sections it includes.
Chief Concern & Goals
One paragraph naming the most important data and AI problem you face, plus the specific outcomes this work must produce. Agreeing on this first is how we avoid solving the wrong problem well.
Data-AI Strategy Vision
A concise five-year statement of where data and AI take the organization, written to be ambitious enough to shift how people think and specific enough to guide real decisions.
Synthesis of Organizational Themes
What we heard in the interviews, organized into the patterns behind it: the recurring issues across people, process, and platform.
Adoption Model Assessment
Your scored position on an adoption model adapted to your sector, with the rationale behind each score and a summary figure you can present to leadership.
North Star Use Case
The unifying use case, written out, with the 5–10 core data elements it depends on and the reasoning behind it.
Gaps
Concrete gaps in data content, quality, acquisition, integration, governance, infrastructure, and cross-functional execution, each tied to the outcome it is blocking.
Prioritized Recommendations, Tactical Steps, and Spot Training
Specific steps: what, by whom, in what timeframe, ranked by impact and feasibility, each traced back to the North Star. They span technology, organization, process, projects, and skills. Alongside the multi-year recommendations we surface tactical moves you can make in the first 90 days, and deliver two to three short, hands-on AI sessions built around the real workflows we encounter during interviews: drafting from prior documents, plain-language summaries, training materials, reviewing narrative reports. The recommendations address the next several years. The spot training is meant to be useful immediately.
Where this has been applied
We have run this process across healthcare delivery, clinical research, and mission-driven organizations. Two anonymized examples:
Governance first, then infrastructure
Growing partner data-sharing demands and new systems had produced wide variation in data practice. Thirteen stakeholder interviews, from CEO to data specialist, surfaced consistent themes around trust, access, and ownership. We built a sector-specific adoption model, scored the organization against it with the project team, and found a top-heavy pattern: strong analytic talent resting on thin governance and infrastructure. The North Star tied fundraising, public investment, and program outcomes into a single sentence. Recommendations ran from appointing a data leader and defining core data elements through building enterprise data management and expanding self-service.
A North Star with its core data elements mapped
The engagement produced a North Star use case with its core data elements mapped out beneath it, an adoption assessment, and a prioritized set of recommendations. We closed by polling stakeholders on how well the strategy resonated and where they wanted investment focused.
Pragmatic, actionable strategy
Grounded in your reality
Strategic advice that ignores your actual people, systems, and budget is common. We have run data operations, built platforms, and carried a P&L. Our recommendations account for where you actually are.
Executive partnership
We know the pressure that keeps executives from strategic thinking. Part of this work is being a trusted sounding board and turning an executive's instinct into direction the organization can act on.
Listening first
Success with technology requires careful listening. We take time to understand the current state: the details of the technical environment and the experience of the people working in it.
Let's talk about your data and AI strategy
Every engagement starts with a conversation about what you are trying to accomplish with data and AI and what is getting in the way. We will tell you honestly whether this engagement is the right fit.
Start the conversationStrategy work often pairs with our privacy-first AI training for the teams who will carry the roadmap. Implementation work (AI development, role definition, hiring, chartering and mentoring projects) is scoped separately.