• Case Study — Financial Services
Turning a Stalled AI Pilot Into a Working Content Workflow
An AI content workflow only works when humans still own judgment, compliance, and final review.A mid-size financial services team invested in an AI writing tool, but adoption flatlined after three months. I mapped the existing content process, identified where the tool was miscast, and redesigned the brief‑to‑publish workflow around it. Within eight weeks, output doubled and the team stopped treating the tool as a threat.
THE CLIENT
A mid-size financial services firm publishing high-regulation thought leadership and updates.
DELIVERABLE
A redesigned brief-to-publish content workflow, standardised briefing frames, and human-in-the-loop guidelines.
• ENGAGEMENT OVERVIEW
A mid-size financial services team invested in an AI writing tool to support thought leadership, client updates and internal explainers. After three months, adoption stalled. A few early champions kept experimenting, but most writers quietly returned to the old way of working. People blamed the tool. In reality, no one had ever designed the workflow around it.
The team did not need another platform demo. It needed a content workflow that made AI feel like support, not surveillance.
• OBJECTIVE
Identify where the AI writing tool was being miscast, map the real content flow, and turn a stalled pilot into a high-performing, human-led publishing sequence.
• PROBLEM
Too many expectations, too little structure
The AI pilot launched with good intentions and vague instructions.
- “Use the tool when it helps.”
- “Try it for first drafts.”
- “See what it can do for you.”
In practice:
- Briefs were long, inconsistent and often arrived late.
- Writers used the tool on highly regulated content where nuance and compliance mattered more than speed.
- No one knew which parts of the process AI should handle and which parts had to stay strictly human.
- Managers worried about “AI tone”, and writers feared being judged against machine output.
The result was predictable: adoption flatlined, trust eroded, and the investment was quietly classified as “an experiment that didn’t quite land”.
• APPROACH
Redesigning the workflow, not the tool
Rather than forcing writers to adapt to the software, we rebuilt the publishing workflow so the technology served specific, high-leverage steps.
01. Map the real content process
Instead of starting with tool features, the engagement began with mapping how work actually flowed:
- How content briefs were originally initiated and written.
- The exact paths of collaboration between writers and Subject Matter Experts (SMEs).
- The number of manual compliance and regulatory review rounds.
- How drafts finally cleared final sign-off and moved to publish.
This made it clear that the bottlenecks were not just about writing speed. They were about unclear ownership, late decisions and repeated rewrites.
02. Identify where AI was miscast
We closely examined how the existing pilot had been operated:
- The tool was being asked to write fully-formed regulated content directly from poor inputs.
- AI was being introduced too late in the process, as a shortcut to fix weak drafts.
- Writers were spending hours manually fixing machine output to match the firm’s tone.
- Compliance reviewers, feeling blindsided by generic text, became significantly more cautious.
In short, the AI tool had been cast as a replacement writer in a workflow that had never made space for it.
03. Redesign the workflow around AI
We positioned the tool as a structured assistant, introducing clear boundaries:
- Standardised structured briefs to ensure the tool always starts with solid, compliant facts.
- Moved AI usage upstream to outline generation, brain-mapping, and structure formatting.
- Established clear ‘human review’ checkpoints where drafts must be approved by senior writers.
- Mapped specific channels so complex, high-risk publications stayed fully human-crafted.
Working rule: AI could suggest language and structure; humans owned judgment, nuance and sign-off.
• Outcomes
What changed
2× content output
The team doubled usable content volume without increasing headcount, largely by shortening briefing, reducing rework and using AI for outlines and repurposing rather than full rewrites.
8-week turnaround
The pilot moved from “stalled tool” to a functioning content workflow in two months, with writers and managers agreeing on when and how AI should be used.
Workflow redesign
The firm gained an explicit, documented brief-to-publish workflow that treated AI as a capability inside the process, not a black-box shortcut sitting on top of it.
Ready to redesign how your team works with AI?