AI compliance monitoring that cut 17 hours a week from a medical team's workload
ClientMid-size pharma company
Location
IndustryPharmaceutical Manufacturing

About the project
We built an AI system that finds online content mentioning a pharma company, scores it for compliance risk, and drafts the response
The client is a mid-size pharma company operating in the United Kingdom. Its medical team was spending a significant share of the week responding to compliance complaints about content the company had neither written nor controlled.
We delivered this as part of our GenAI solutions for pharma offering.
The Challenges
In the UK, pharmaceutical promotional activity is governed by an industry code, and the regulator investigates complaints against it.
Complaints can be raised about content that mentions a company or its products even when the company did not create that content. Sponsored events, third-party medical platforms, and patient group websites all produce material of this kind, which left the client accountable for material it had never written.
- 01
Accountability for content the company did not control
- 02
A tight response deadline with potential fines
- 03
Manual review absorbing the medical team's week
- 04
No visibility into what was being published
- 05
Specialist time displaced from higher-value work
The Objectives
The client wanted to stop reacting to compliance complaints one at a time and start seeing problems before the regulator did.
02
Separate genuinely relevant mentions from unrelated results, so the medical team is not reviewing noise.
03
Score every mention for compliance risk against the industry code, rather than treating all mentions equally.
04
Hand the team a recommended course of action for each case.
05
Produce a draft official response to the regulator that a human only needs to check and approve.
Our Approach & Solutions
We built the system around a simple division of labor. Software handles the volume, finding and sorting content at a scale no team could match manually. A person handles the judgment, reviews what the system surfaces, and approves the response.
- 01
Custom scrapers across UK web sources
- 02
Relevance filtering to cut out the noise
- 03
An AI agent trained on the compliance code and internal precedent
- 04
Three-tier risk classification
- 05
A report with an action plan attached
- 06
Pre-drafted regulator responses and third-party outreach
- 07
Human review kept firmly in the loop
The Results
The system gave the client visibility into what the internet says about the company before the regulator did. The operational gain was immediate, and the reduction in financial exposure was larger still.
Key outcomes included:
- 01
An estimated 17 hours per week returned to the medical team
- 02
Close to £1 million in potential fines avoided
- 03
The medical team moved back to higher-value work
- 04
Leadership wanted to take it further
Highlights
17 hours
saved per week
~£1M
in potential fines avoided
3 risk tiers
automatic classification
Before this system, compliance work landed on us as a series of unpleasant surprises. Now we see the content first, we know how serious it is, and most of the response is already written by the time it reaches us. It changed what my team spends its week doing.
&w=3840&q=75)
Digital Project Manager
Mid-size UK pharma company
Before this system, compliance work landed on us as a series of unpleasant surprises. Now we see the content first, we know how serious it is, and most of the response is already written by the time it reaches us. It changed what my team spends its week doing.
&w=3840&q=75)
Digital Project Manager
Mid-size UK pharma company

The compliance rules in this market are public and well documented, which is exactly what makes them a good fit for AI. The real problem was never interpreting the code. It was finding the content in the first place and deciding which mentions genuinely mattered. Once we solved that, the rest followed, and the client's team got their week back.
&w=3840&q=75)
Alex Jijie
CEO, Digitalya
The compliance rules in this market are public and well documented, which is exactly what makes them a good fit for AI. The real problem was never interpreting the code. It was finding the content in the first place and deciding which mentions genuinely mattered. Once we solved that, the rest followed, and the client's team got their week back.
&w=3840&q=75)
Alex Jijie
CEO, Digitalya
Component could not be found for blok section-project-mockups! Is it configured correctly?