Written by Most organisations blame “too much governance” for slow AI approvals, but the real bottleneck usually forms earlier, at the intake stage, where inconsistent submissions create the pile-up that governance committees later get blamed for clearing.
What Is AI Use Case Intake
AI use case intake is the structured process by which an organisation captures, standardises, and logs proposed AI applications before they are evaluated for feasibility, risk, or business value. Defining this clearly matters because most failures in AI governance programs trace back to inconsistent intake rather than an overly strict approval stage.
Why Unstructured Intake Fails
Teams submit ten-page business cases for simple automation ideas, executives raise ideas verbally that skip the queue entirely, and business units invent their own local version of a “request form” because no shared one exists. This creates two problems at once: genuinely valuable small ideas get buried under process weight, while politically connected ideas skip the queue regardless of merit.
Building A Minimal Viable Intake Form
A standardised one-page intake form covering four fields — problem statement, current process pain, data types involved, and expected benefit — keeps submissions fast enough that non-technical staff will actually use it. However, real-world implementation shows this form needs to evolve quickly: CISO Billy Norwood, describing his organisation’s AI governance rollout across 40 initial use cases, admitted their first intake attempt was thin, essentially asking submitters to self-report technical effort and risk without real structure, and it had to be reworked almost with every new use case as gaps kept surfacing. This is a useful expectation-setter for any organisation starting out: build your intake form expecting to revise it constantly in year one, not to get it right on the first attempt.
The Friction Question
One counterintuitive lesson from real deployments is that frictionless intake can backfire. When asked how to prevent people circumventing the process, Norwood’s answer was to deliberately insert friction by controlling platform access — in his case, funnelling all AI activity through a single sanctioned platform (Databricks) so there was no easy path around the intake process. The lesson here is nuanced: keep the form itself fast, but make the sanctioned pathway the only realistic route to production, rather than assuming speed alone will stop people going around governance.
Capturing Risk At The Point Of Entry
Even at intake, capture whether the use case touches personal data, decision-automation, or vulnerable populations, since flagging these early prevents rework at the scoring stage later. This turns intake from a pure operations fix into a governance control without slowing submission speed.cdomagazine+1
FAQ
What is the difference between AI intake and AI approval?
Intake captures the idea in a standard format; approval is the later decision on whether to proceed.
How long should an intake form take to complete?
Aim for under 15 minutes with four to six required fields, but expect to revise the form frequently in your first year as gaps emerge.
Who owns the intake process?
Typically a central AI or data governance office, not individual business units.
How do you stop people bypassing the intake form?
Restrict access to sanctioned AI platforms so the intake-approved path is the only realistic way to reach production tools.
Practical Takeaway
Publish one shared intake form this week, retire all informal request channels, and commit to logging every submission in a single backlog regardless of size — then plan to revise the form monthly for the first two quarters.
If you’d like assistance or advice with your Data Governance implementation, or any other topic (Privacy, Cybersecurity, Ethics, AI and Product Management) please feel free to drop me an email here and I will endeavour to get back to you as soon as possible. Alternatively, you can reach out to me on LinkedIn and I will get back to you within the same day!
Nigel D'Souza 