Decked Out: How Redscout Ditched Slides for Strategy That Runs

Interview by Bernard Desarnauts
Estimated reading time: 7 minutes

Watch the entire interview here

For years, 30 % of Redscout’s revenue came from meticulous research decks. Then clients opened ChatGPT, asked a few market questions, and discovered they could fetch “pretty good” answers in seconds. Ivan Kayser, CEO of the 40-person brand consultancy, knew the billable model was finished: “Clients didn’t want to pay for that anymore.” Rather than discount rates, he dropped the agency AI tools he’d been trial-running, built project-specific GPTs on Redscout’s own work, and reframed strategy as runnable code, something competitors can’t copy with an off-the-shelf app. The shift shows every 10- to 100-employee firm how to trade average deliverables for a proprietary edge.

6 Reality-Check Insights

1. Research as a safety net is gone. Up to 30 % of revenue was tied to work AI now automates.
   

2. SaaS AI is a “thin UI layer.” Kayser: Tools pushed users into average workflows.
   

3. Custom data beats generic every time. Specific prompts + internal corpus changed outputs instantly.
   

4. Documentation is the new moat. Years of unlabeled files blocked progress until systematically tagged.
   

5. Strategy must become code, not slides. Executable logic replaces 2 × 2 frameworks.
   

6. Mid-career staff resist most. Interns and execs lean in; the “middle layer” needs coaching.
   

Premium Research Turned Commodity

“Many firms were doing 80 % research, 20 % strategy,” Kayser recalls. Even strategy-led shops like Redscout still collected “30 % of their revenue maybe on research.” Once reasoning models hit the market, that work felt performative: “You did the research in order to justify the reasoning but you could have reason-justified anything.” Clients now ran their own sentiment scans, so research hours no longer anchored project budgets.

Four First-Principle Questions

Instead of patching slides with a “powered by AI” badge, Kayser gathered the team and asked:

1. What do clients find genuinely hard?
   
   

2. Where do we create undeniable value?
   
   

3. Which tasks burn time yet add little value?
   
   

4. How can AI amplify the first two and delete the third?
   
   

The answers ruled out one-size SaaS platforms. Redscout needed something it could own, extend, and improve with every engagement.

Three Moves That Turned Slides Into Software

Move 1 – Skip the Marketplace
Agency-focused AI apps promised drag-and-drop surveys and idea boards. All failed the reality test: average workflows, generic data, slow to adopt new models. “You could get a better research output directly into a reasoning engine,” Kayser realized. The firm now spins up custom GPTs for each project, seeded with client material and Redscout’s own playbooks.

Move 2 – Document Like a Product Team
“We realized how bad we were at documentation.” The team spent weeks tagging a decade of decks: client context, problem, solution, impact. A lean but precise corpus now feeds every model. No competitor can scrape that data.

Move 3 – Make Strategy Runnable
Redscout distilled brand work into three callable functions: Diagnosis → Insight → Unlock. When strategists enter a prompt, a custom GPT digests tagged examples, surfaces patterns, and critiques or generates copy. Kayser frames it simply: “Strategy is going to become code much more than framework.”

Results

From Average to Precise

• Research hours fell sharply, freeing time for deeper unlocks.
  

• Turnaround on first-round insights dropped from weeks to hours.
  

• Prospects now see live, project-specific thinking in the first call, lifting win-rates.
  

• Every new engagement enriches the corpus, compounding advantage.
  

Kayser’s verdict: “AI is built to respond to an individual prompt and has no problem with complexity. The tool adapts; the client doesn’t.”

Action Take-Aways

1. Ask the Four Questions First
Map value, waste, and leverage before installing any tool.

2. Tag Your Best Fifty Files
A week of disciplined labeling outperforms terabytes of random docs.

3. Build One Micro-GPT This Week
Pick a repetitive task, train on your examples, and test on live work within 48 hours.

4. Coach the Middle Layer
Pair skeptical mid-career staff with early adopters; celebrate 15-minute wins.

5. Think Systems, Not Tools
Treat AI like a buildable capability—skip generic SaaS, own your data, own your logic

Want to see more?