Quality teams closed records faster while leaders gained greater control over AI use.
A pharmaceutical manufacturer needed to reduce documentation delays and regain oversight of AI use after a company-wide freeze. We helped turn a blanket restriction into practical review paths, vendor controls, and decisions that stayed with qualified client leaders.
business days: typical closure time
From investigation completion to deviation-record closure.
Why it matters
The manufacturer reduced documentation delays and unapproved AI use while keeping quality decisions with qualified people. Leaders gained a record of AI approvals, controls, and accountability that stood ready for a licensing partner’s quality audit.
The starting point
A company-wide ban had not stopped all AI use.
An uncontrolled consumer-chatbot draft had entered a deviation workflow. Quality leadership froze AI use across the organization, including general business functions. Network logs still detected an average of 41 consumer-AI sessions per week.
Quality teams also faced a backlog of 68 open deviation records. Median time from investigation completion to record closure stood at twelve business days. The manufacturer needed a way to distinguish applications by risk and define the evidence, review, and authority each required.
What the business needed
Reduce documentation delays, bring AI uses into a visible review process, and keep quality and regulatory decisions with qualified people.
What we changed
We turned the broad restriction into explicit decisions, controls, and review responsibilities.
Separate uses by risk and review requirements
Written classification rules and an acceptable-use policy created distinct intake paths for uses adjacent to regulated quality workflows and general business applications. Each proposal could receive an explicit decision instead of falling under the same blanket ban.
Make vendor controls a condition of selection
Vendor evaluation became part of supplier qualification. Two vendors failed the prompt-retention review and were rejected. The selected platform went through qualification with audit-trail and prompt-retention controls.
Set firm limits on what AI could decide
AI-assisted drafts in regulated quality workflows required qualified human authorship and review. Phase one excluded AI from batch-release and quality-control decisions. Client quality and regulatory leaders retained judgment and approval authority.
What the client could use next
- Written rules for evaluating AI use
- Classification rules and an acceptable-use policy that distinguished applications by risk and made review requirements explicit.
- Intake and vendor-review processes
- Separate paths for quality-related and general-business uses, with vendor checks incorporated into supplier qualification.
- An AI use register and review cadence
- A register that became a standing item in monthly quality-management review, alongside documented human authorship and review requirements.
What the evidence showed
Fewer consumer-AI sessions appeared in network logs.
Detected consumer-AI sessions per week
- At baseline
- 41
- After the changes
- 6
Both figures use the same measurement method. These counts describe detected sessions, not people or data breaches. They do not establish that all unapproved use stopped.
Fourteen use cases received explicit decisions
In the first quarter, 14 use cases entered intake: eight were approved, four were returned with conditions, and two were declined in writing. Approval did not itself establish that an application had entered production.
Document closure became faster
Median time from investigation completion to deviation-record closure fell from twelve business days to seven.
Audit evidence was ready in the course of normal work
Governance evidence was available during a licensing-partner quality audit without special preparation. The AI use register also became part of regular quality-management review.
Faster documentation did not transfer decision authority to AI.
The closure measure begins after investigation completion; it does not cover the full investigation. Available audit evidence does not establish certification or regulatory endorsement. Client leaders retained legal, quality, regulatory, and product-release authority; Imajin Labs did not provide those approvals.
Imajin Labs gave us a workable path out of the AI freeze. We could see what needed review, what evidence was missing, and who had authority to decide. Useful work could move forward while our quality and regulatory leaders retained control.