AI TRANSFORMATION
One Person, Two Names: Why AI Transformations Stall on Identity Drift
Siloed data drifts apart over the years. AI joins records constantly. Measure the drift before AI is trusted with the records.
How many versions of each customer, citizen or employee does your organization hold, and which one is right?
We enable an organization to answer that with evidence, before AI is trusted with its records.
What Identity Drift Is
Many AI transformations stall on something that has nothing to do with the model: identity drift, the unintended consequence of years of siloed data.
A large part of transformation is moving from separate systems, each serving its own purpose, to services that AI assistants can run across them: one view of the customer, eligibility decided in seconds, an assistant that can answer for the whole organization. Each of those depends on the systems agreeing about who and what they hold.
Identity drift is why they often don't. Nobody intended it, and nobody notices until something tries to join the records.
How It Happens
Each silo was built to do its own job well. Over the years the versions of the same person slowly move apart:
- A name changes in one system and not in the next.
- A correction lands here and not there.
- A new address reaches one record and not the other.
- Two identifiers are issued for one person, one in each system, and nothing links them.
- A migration carries the old version forward beside the new one, and both survive.
One government found this when it combined two identity cards into one. Renewals stalled on names that no longer matched between the two registers that issued them.
Why AI Makes It Urgent
People work around drift every day without noticing; they recognize the person behind the two records. AI assistants join records constantly, and they do not recognize anyone:
- A customer view assembled from three systems shows three people.
- An eligibility decision made in seconds is made on the wrong version.
- An assistant answers for the whole organization from whichever record it found first.
- A renewal stalls because the name on the card no longer matches the name on file.
This is why so many AI pilots succeed and then stall at scale. The pilot ran on one clean system. The transformation has to run on all of them.
If you know your records disagree, you don't trust them. If you don't know whether they agree, you can't trust them. An AI assistant trusts them anyway.
Measure the Drift First
A transformation plan that expects identity drift builds the data work in before any model is chosen:
- Find: Every system that holds the same people, properties or accounts, including the spreadsheets and the file shares nobody lists.
- Measure: How many versions of each person the organization holds, how far apart they are, and how many cannot be matched at all.
- Decide: Which source is authoritative for each attribute, and keep the history and evidence behind every change.
- Match: Across systems in a way a person can audit, before anything automated relies on the match.
- Fix: Resolve the drift in order of risk, and record who approved each change.
The measurement is the part most plans skip. Without it, the drift is discovered by the AI assistant, in front of a customer.
How CTS Resolves It
The AITS Content Tracking System (CTS) is the tool for this work. The CTS score is a composite index of KPIs that tells the health of a document repository. Identity drift shows up in four of them:
- Duplication: The same person, property or account held more than once, with the authoritative copy identified.
- Currency: The current version told apart from superseded ones, so the old name is not the one an assistant finds first.
- Metadata quality: Identifiers, names and dates present, valid and consistent across systems and eras.
- Completeness: Records that do not link: the person in one system with no counterpart in the next.
CTS runs on a secure replica hosted in-country. The official systems are never changed. It measures the drift across every system it is pointed at, identifies the authoritative copy, and recommends action on each finding with the evidence behind it. Staff confirm, correct or dismiss each one, and the AI used by CTS learns from their decisions, so the next pass finds the next issue sooner.
The result is one score an executive can read at a glance, and the breakdown that tells staff where the work is. See how CTS works.
Keep Measuring
Drift returns the day after it is fixed, because the systems that caused it are still running. Continuous monitoring keeps the records agreeing after go-live:
- Monitor: Content is watched as it is added, changed, migrated or corrected in any of the systems.
- Re-score: The health score is recalculated against the organization's KPIs and weights.
- Alert: An alert is raised whenever drift returns above the threshold the organization set.
- Recommend: CTS recommends action on each finding. Fixes of a type the organization has approved are automatic; everything else goes to staff.
So that when AI assistants are used on those records, they do not give confident answers that are wrong.
Find out how far your records have drifted
A Repository Health Check gives an executive a scored baseline for their own content, the detail behind it, and a prioritized list of what to do next.
Contact AITS to book a Repository Health Check info@appliedits.comON THE AITS WEBSITE
Content Tracking System (CTS): The tool that measures and resolves identity drift.
Article: One Number for the Health of Your Document Repository: What the CTS score is and what is behind it.
Repository Health Self-Assessment: Sixteen questions and an estimate of your own score.
Sample Health Report: What an executive receives.
Solution Brief: Is Your Repository Ready for AI?: The Repository Health Check in two pages.
ABOUT AITS
Applied IT Solutions (AITS) is a Toronto-based consultancy founded in 2008 and led by its principal consultant. It advises government and Fortune 500 organizations on artificial intelligence, enterprise data strategy, document and records management, and business process improvement, with one aim: deriving tangible business value from the application of information technology. Most engagements are delivered end to end by the principal, from requirements gathering through solution design and delivery to train-the-trainer and hand-off to the client's own team. For larger projects, AITS brings in experienced consultants it works with regularly. AITS delivers globally. Out of this work, AITS developed the Content Tracking System (CTS), an AI-powered monitor that scores the health of a document repository against the organization's own KPIs.