One of the most common mistakes I see leaders make with AI is talking about impact in averages:
“AI will 10x productivity”
“AI will automate 30% of work”
“AI will transform the whole company”
That framing sounds powerful, but it's simply not correct.
AI gains are not evenly distributed across a business.
They vary by function, by role, and by type of work.
Thanks for reading,
Robbie Allen
Founder & Managing Director
Automated Consulting Group
PS: If you’re struggling to forecast ROI from AI investments, hit "Reply."
Key Takeaways:
AI productivity gains are highly uneven across functions. Talking in company-wide averages leads to bad planning.
ROI improves when AI budgets are allocated by function, not hype.
“AI didn’t work” is often a modeling failure, not a technology failure.
Why Ops Won't See the Same AI Gains as Engineering
From the work we’ve done across companies and my own experience, the real distribution looks roughly like this:
Engineering work: up to 10x productivity gains
Knowledge-heavy consulting and analysis: closer to 5x
Many operational and support activities: around 1.5x
Some work: effectively no material gain
If you assume AI will deliver the same lift everywhere, you'll end up over-investing in low-return areas while removing budget from areas that could see higher gains.
In the end, you'll conclude that AI failed when, in reality, you averaged the results away.
Averages are comforting because they simplify planning, but they’re the wrong abstraction for AI.
AI doesn’t behave like past enterprise software, where value came from broad standardization. It behaves more like leverage: it amplifies certain types of work and barely touches others.
Why Org-Wide Is the Wrong Level for Assessing AI's Impact
When leaders talk about “AI impact” at the company level, three bad things happen:
1. ROI forecasts are wrong
High-performing teams subsidize low-performing use cases.
2. Workforce planning gets distorted
Leaders expect uniform efficiency gains that never materialize.
3. Postmortems miss the real lesson
The takeaway becomes “AI didn’t work” instead of “we applied it in the wrong places.”
This is why so many AI initiatives stall after early pilots.
A Practical Framework: Where Will AI Actually Move the Needle?
Instead of asking, “How will AI impact our company?” ask this, function by function:
1. Is the work judgment-heavy or execution-heavy?
AI excels at accelerating execution, not replacing judgment.
Strengths: coding, analysis, drafting, synthesis
Weaknesses: nuanced human decision-making
2. Is the output already digital?
AI compounds digital work, but it's not ready to take partners out to dinner and understand the root cause of issues.
Strengths: software, documents, data, workflows
Weaknesses: physical or relationship-driven tasks
3. Is there repetition with variation?
This is the sweet spot.
Strengths: similar problems, different inputs
Weaknesses: one-off, bespoke work
4. Does speed meaningfully change outcomes?
If faster execution creates value, AI pays off. But if speed doesn’t matter, gains are limited.
When teams walk through this framework honestly, patterns emerge very quickly.
Budgeting AI by Function, Not in Bulk
One of the most effective shifts we see is when leaders stop allocating AI budget “across the enterprise” and start allocating it by function.
That means:
Doubling down where 5–10x gains are realistic
Being modest where 1–2x is the ceiling
Explicitly accepting that some areas won’t benefit much (yet)
This doesn’t mean ignoring those teams. It means setting realistic expectations and avoiding false negatives.
The Takeaway for Leaders
AI is not a blanket productivity multiplier, but a targeted leverage tool.
Leaders who get the most value stop asking:
“How much will AI improve our company?”
And start asking:
“Where will AI actually move the needle and materially change how work gets done?”
And that's when ROI forecasting becomes more accurate.
– Robbie
P.S. I’m curious. Hit 'Reply' and tell me which part of your company is currently seeing the highest gains from AI.