Six weeks. Your real data, graded. One decision.
Before you fund an AI use case, or let it go live, we profile real extracts from every system it depends on, grade each one A–F on five dimensions, and finish with a written verdict: go, fix then go, or no-go. For P&C insurers and health plans.
When to Call Us
What we hear before a pilot.
Sign 01
The board wants AI, but our data isn't ready.
An AI mandate with a clock
The board or CEO has set expectations, and someone owns a deadline they aren't sure they can hit.
Sign 02
We tried a vendor POC and quietly shelved it.
A pilot that never shipped
A vendor proof of concept or an internal build never made it to production, and nobody agrees why.
Sign 03
Nobody trusts our claims data.
A data estate nobody trusts
Claims data, policy records, or lineage questions keep surfacing in every planning meeting.
Sign 04
A vendor is ready to start, and nobody has checked the data.
A vendor ready to start
A vendor has a statement of work ready, and nobody has checked whether the data can support it.
How It Works
Six weeks from data access.
We need interviews with the people who know your systems and read access to extracts from the systems in scope. The clock starts when data access is in place.
Interviews & mapping
- Interviews with data, IT, and business leads
- Every in-scope source system mapped
- The AI use case you're being asked to deliver, documented
Why: So we grade only the systems your use case touches, not your whole data estate.
Profiling & scoring
- Real extracts profiled, not just people surveyed
- Every system graded A–F on five dimensions
- Gaps ranked by impact on that use case
Why: So every grade rests on evidence from the data itself, not on opinion.
Leadership read-out
- Written verdict: go, fix then go, or no-go
- Findings and roadmap presented to leadership
- First use case scoped for 90 days
Why: So leadership can make the call with evidence, not opinions.
Profiled, not surveyed.
Interviews tell you what people believe about the data. Profiling shows what it actually is. Most readiness surveys end with a maturity score from a questionnaire. The Verdict ends with a grade for each system, and the extract that earned it.
Why these five dimensions
Each one is a different way a pilot fails in production. We grade every system on all five.
Data Quality
Models learn whatever errors the data contains, at scale.
Duplicate policy records for 6% of insureds since the last platform migration.
Documentation
If no one can say what a field means, no one can map it to a model.
No data dictionary. Field meanings are known only to the original vendor team.
Lineage
Regulators and auditors will ask where every input came from.
Paid-loss figures in the board pack can't be traced past the warehouse load.
Access & Security
AI teams that wait months for data never get past the pilot.
Access granted request by request through one DBA; average wait three weeks.
AI Consumability
A model can only use data it can read at the moment it decides.
Loss descriptions arrive as free text and scanned PDFs. A triage model has nothing structured to read at first notice.
Examples from the sample read-out for Meridian Casualty Group (fictional).
What We Provide
A decision, with the evidence attached.
Go
Fund the pilot as planned.
Fix, then go
Fund these fixes first, in this order.
No-go
Keep the budget. Here's why, in writing.
Sample Deliverable
What the Report Card looks like.
An excerpt from a sample read-out for a fictional P&C carrier. Illustrative, not a client result. Every grade follows our published grading rubric.
Swipe to see all five grades →
| System | Data Quality | Documentation | Lineage | Access & Security | AI Consumability | Overall |
|---|---|---|---|---|---|---|
| ClaimsCore | B | D | D | B | D | C |
| PolicyCore | D | D | D | C | D | D |
| FNOL Intake | C | F | D | C | F | D |
| ACORD Gateway | B | D | C | B | C | B- |
| EDW | C | C | D | B | C | C+ |
The circled F is where an FNOL triage pilot dies. Every grade is evidence-backed: interviews plus profiling on real extracts, tied to the use case leadership wants delivered.
The roadmap: four horizons.
0–90 days
Quick wins
- Glossary adopted, data stewards named
- First AI use case scoped with a written accuracy target
- Remediation backlog funded by leadership
Why: Unblock the first use case fast, and show progress early.
Months 3–6
Fix what blocks the use case
- Claimant and policy golden-record pilot
- Claim documents made machine-readable
- Pipeline fixes for every system that scored D
Why: Remove the specific gaps that would block the use case in production.
Months 6–12
Rebuild the plumbing
- Automated lineage across core systems
- Pipeline rebuilds and system integration
- Adjuster notes and estimates structured
Why: So the second and third use cases don't start from zero.
12 months +
Hand over
- Same-day data for underwriting and claims
- Playbooks handed to your own team
- Capacity that scales with the AI program
Why: So your team owns it, and you're not dependent on us.
$75K
from · fixed fee
Priced before we start
From $75K for up to five source systems. Ten and fifteen systems are scoped the same way. The price is set in the scope, with no hourly billing.
1
principal on every grade
Principal-led
The principal designs the method, runs the key interviews, sets every grade and presents the read-out. Specialist support is brought in as the scope needs it.
0
platforms to sell
Independent, and yours
We don't sell AI platforms. The verdict is written and delivered before any follow-on work is discussed, and it's yours to take to any firm.
After the Read-out
A six-week answer, then a sequenced path.
If you want us to keep going, two engagements follow. From data access to first use case in production: about 18 to 20 weeks.
Applied AI Build
8–12 weeksOne narrow use case, built to a written accuracy target, inside the 90 days after the read-out.
Why: Prove value on one use case before scaling.
AI Governance
4–8 weeks, alongside the buildThe controls that let a model go live in a regulated business.
Why: So risk and compliance can sign off, and the model isn't pulled after launch.
Before You Sign
Questions security and procurement ask.
Straight answers on data access, your team's time, cost and independence. For your security reviewer, everything about data handling is on one page, with a PDF.
Data access and security
What data do you need?
Read access to extracts from each in-scope system, plus the documentation you already have: data dictionaries, architecture diagrams, interface specs. We profile samples; full production loads aren't required.
Does our data leave our environment?
Only if you choose. By default, profiling runs inside your environment, under your access controls. Where that isn't practical, extracts move over a secure transfer method you approve, and are deleted at the end of the engagement.
Will you sign our NDA and go through vendor security review?
Yes. We sign your NDA and complete your vendor security review before data access, and plan the start date around it.
What about claimant, member and patient data?
Profiling measures fields and tables: completeness, validity, duplicates, and how records link across systems. Nobody reviews individual claims or patient cases. Where sensitive fields such as PHI are in scope, they can be masked before profiling, or profiled inside your environment so they never leave it.
What about our regulators?
Regulators ask where reported figures and model inputs come from, who owns the data, and who can reach it. The Verdict grades exactly that, system by system, and shows where the answers break down. Closing those gaps is roadmap work, and governance support, including deployment patterns aligned with the NAIC Model Bulletin for insurers, is available as a follow-on.
Time, cost and scope
How long does it take?
Six weeks from data access: weeks 1–3 interviews and mapping, weeks 3–5 profiling and grading, week 6 the written verdict and leadership read-out. The clock starts when access is in place, not at signature.
How much of our team's time does it need?
A project lead to schedule interviews and approve access; 12 to 20 interviews of about an hour each in weeks 1–3, with the data, IT and business people who run the in-scope systems; and your sponsor at the read-out.
What does it cost?
From $75K for up to five source systems. Ten and fifteen systems are scoped the same way. The price is set in the scope, before we start, with no hourly billing.
What if we have more than 15 systems?
We scope to the systems your AI use case actually depends on, which is usually fewer than people expect. Larger estates can be graded in phases.
Is it remote or on-site?
Remote. Interviews, profiling and the read-out run over video and the access you approve. If you'd like us in the room for the read-out, we can arrange it.
Do you work with health plans as well as insurers?
Yes. Health plans are graded on the same five dimensions, against use cases such as fraud, waste and abuse detection. Switch this site to Health plans to see a health plan's sample read-out.
Outcomes and independence
What do we get at the end?
A written verdict first: go, fix then go, or no-go, on one page. Behind it, the Data Foundation Report Card grading every in-scope system A–F on five dimensions, a fix list ranked by impact on your use case, a roadmap in four horizons with the first use case scoped for 90 days, and a leadership read-out.
Who does the work?
Principal-led, with specialist support as needed. A principal architect with more than 20 years building data platforms in insurance and healthcare designs the method, runs the key interviews, profiles the data, sets every grade and presents the read-out. Specialists join for interviews and coordination when the scope calls for it. There's no hand-off to a junior team.
Do we have to hire you to implement it?
No. We don't sell AI platforms. The verdict is written and delivered before any follow-on work is discussed, and it's yours to take to any firm. Your own team or existing partners can take it on. If you want us, we deliver the first AI use case (Applied AI Build) and the governance work; larger platform engineering is delivered with our partners. Any commercial relationship with a partner is disclosed in writing before the roadmap is delivered.
Are your recommendations vendor-neutral?
Yes. The roadmap points to what your use case needs, not to a particular product.
What if the answer is no-go?
Then you've kept the pilot budget, and you know exactly why, in writing. The fix list says what would change the answer. That's the point of doing it first.
Six weeks from data access, you could have the answer.
Tell us the AI use case you're being asked to deliver and roughly how many source systems it depends on, and we'll send a scope and a fixed fee. From $75K for up to five source systems.
We take on a limited number of engagements at a time, so each one gets principal attention.




