Data quality for AI, by systemEvery system looks clean on its own. The truth is in the gaps between them.
Every AI proposal assumes your data is fine. Nobody has checked. Your CRM, your ledger and your HR system were each built to be right about one thing; none was built to agree with the others. These guides set out, system by system, what each one gets right, where its data breaks for AI, and the gaps to test between it and the systems beside it.
Starts with a 30-minute call. Commits you to nothing. Then exports only: no logins, nothing installed.
System guides
Find your system.
CRM
Pipeline, accounts and owners. The system most AI plans start from.
- Salesforce Pipeline, accounts and owners. Usually disagrees with finance on who the customer is.
- HubSpot Contacts, companies and deals. Easy to fill, hard to keep joined to billing.
- Microsoft Dynamics 365 Accounts, opportunities and cases in Dataverse. Modules that rarely agree with each other.
- Pipedrive Deals and activities. Light structure, with key facts living in notes.
- Zoho CRM Leads, contacts, accounts and deals. Configurable to the point of sprawl.
ERP
Orders, items, stock and the ledger for larger and multi-entity businesses.
Accounting
Invoices, payments and the books: where revenue is actually recorded.
HR and payroll
People, roles, pay and leavers: who owns what, and who has gone.
- Workday Workers, positions and organisations. Effective-dated history, unevenly kept.
- BambooHR Employee records for growing firms. Self-service entry, uneven history.
- Personio European HR, absence and contracts. Flexible attributes, uneven structure.
- ADP Payroll and pay codes. Where HR, finance and reality meet, or don’t.
Support
Tickets, conversations and service levels: the earliest warning signs.
Marketing
Leads, profiles, consent and engagement, before anyone is a customer.
Data warehouse
The copy where models train. Copies drift from their sources.
Not listed? If it exports to CSV or Excel, it can be compared with anything else. Tell us on the call.
Why pairs, not systems
The red flags live between two systems.
Each guide includes a map of the systems that one sits beside, and the gaps to test between them. What we find feeds the Truth Score™, a score out of 100 for how far your data can be trusted. Four of the pairings we test first:
CRM ↔ Finance
Closed-won deals with no matching invoice
HR ↔ CRM
Accounts owned by people who have left
Support ↔ Billing
Premium service for customers who stopped paying
Warehouse ↔ Ledger
Board revenue the ledger can’t reproduce
Examples of what we check, not client findings.
Also by
Start from the AI or the industry instead.
What you get back
See what comes back: a real sample memo.
One fictional B2B software company, Example Co: 72,837 records in 7 exports from its CRM, finance and HR systems, run through the same engine and the same checks as a paid Truth Score Read. Each system passes its own checks. Matched across systems, only 13% of deals tie to an invoice.
Questions
Systems and exports, answered plainly.
Anything else, ask on the 30-minute call. It commits you to nothing.
Book the callMy system isn’t listed. Can you still include it?
Yes. The method works for anything that exports to CSV or Excel: practice management, transport and warehouse (TMS, WMS), shop floor and quality (MES, asset registers), store platforms, spreadsheets. These guides cover the systems AI proposals lean on most; they are not a list of what we can handle.
Do you connect to our systems?
No. We work from exports only: one person on your side runs standard exports under NDA, using the field list we send. No logins, no connected apps, nothing installed.
Why compare systems instead of checking each one?
Because every system looks clean on its own. The problems that sink AI sit between them: won deals with no matching invoice, owners HR says have left, customers support can’t place. You only see those by matching records across systems. Our engine does the matching; every red flag is checked by hand against sample records before sign-off.
How many systems do we need to send?
Two is enough to start; that is where the first cross-system red flags show. Add more when the decision depends on them.
What happens to our exports?
They stay under the NDA and data processing agreement you sign first. Raw exports are deleted after the engagement; we keep quality metrics, not your records. The full list of sub-processors comes with our data processing agreement, before any data moves.
Do you put our data into an AI model?
The Truth Score is rule-based: the same data, scored on the same date, gives the same score, and no AI model decides it. We do use a language model (Anthropic’s Claude, through its commercial API) to read the plan or proposal you send and to help draft the memo’s wording from our findings. A person checks every line before it reaches you. If you’d rather we didn’t use it on your engagement, say so and we do that part by hand.
Any two systems in. The gaps between them in writing.
Standard exports, under NDA. A signed memo with your Truth Score within 5 working days of complete exports. Starts with a 30-minute call. Commits you to nothing.
Starts with a 30-minute call. Commits you to nothing.