I was reading a blog post recently from a software developer named Tuhin Nair. He was writing about why senior developers fail to communicate their expertise — specifically, why the people responsible for keeping systems stable keep talking past the people trying to move fast. And I kept thinking: this isn’t just a developer problem. This is exactly what I see happening inside accounting and finance firms every time the topic of AI comes up.
So let me share the framework, because I think it explains a lot.
Both Sides Are Right
Here’s the scene. The CFO or a managing partner says: “We need to move faster on AI.” The IT lead or the risk team says: “We need to slow down and think about this.” Both sides walk out frustrated. The CFO thinks the risk team is obstructing progress. The risk team thinks leadership doesn’t understand the exposure.
Here’s the thing — they’re both right. They are not actually disagreeing about AI. They’re each solving a completely different problem, and they don’t realize it. That’s the whole issue.
The Two-Loop Model
Every firm runs two loops simultaneously.
Loop 1 — The Uncertainty Loop. This is where your CFOs, partners, practice leads, and BD people live. Their job is to reduce uncertainty. They take things to market, get feedback, and learn fast. Speed is the primary weapon. Every delay is another missed opportunity to find out if something works. Their monster is uncertainty.
Loop 2 — The Stability Loop. This is where your controllers, IT leads, risk and compliance officers, and ops teams live. Their job is to manage complexity so the firm can keep serving clients reliably. Every new system, every new workflow, every new integration adds another potential failure point. Their monster is complexity.
Both loops are rational. Both are necessary. The problem is that AI has sent them completely out of sync.
What AI Did to the Two Loops
Before AI, these loops could stay roughly in sync. A new ERP took a year or two. A new reporting tool took six to twelve months. That time forced both loops to move together — you couldn’t outrun your own review process because the technology itself was the bottleneck.
AI changed that completely. Now a junior accountant can build a client-facing output in an afternoon. A practice lead can have a working prototype in 48 hours. The Speed loop can run as fast as it wants. But the Stability loop hasn’t changed at all — the review processes, the documentation standards, the oversight infrastructure is still the same as it was when technology moved slowly.
Firms running at Loop 1 speed without Loop 2 infrastructure are building up technical debt and compliance risk faster than they can manage it. And the people warning about this — the risk team, IT, compliance — are getting ignored because they’re speaking the wrong language.
The Communication Problem
Stability people reach for complexity language: “This adds maintenance burden.” “We need to understand the architecture first.” “The long-term costs will be significant.” All true. None of it lands with someone focused on speed.
Think about how a good engagement review partner operates. They don’t say “I think you got this wrong.” They say “Can you walk me through how you got here?” Same information need. Completely different reception. The stability team needs to do the same thing — translate their concerns into Loop 1 language.
Instead of “this adds complexity,” try: “What’s the minimum we’d need to test this with a real client before we commit?” Instead of “we need more time,” try: “Can we try something quicker first?” Same underlying concern. Framed as a speed question instead of a stop sign.
The Fix: Decouple the Loops
The most practical thing you can do is explicitly run two versions of any AI initiative in parallel — and name them out loud.
The Speed version: fast, rough, good enough to get in front of a client and get feedback. Think of it like a writer’s first draft — you’re not trying to publish it, you’re trying to learn something. Three days, not three months.
The Scale version: reviewed, documented, built to last. This is the one your stability team actually has time to do properly. Six weeks, not six days.
When someone asks you to build something ambitious, try saying: “I can have the Speed version ready in three days. The Scale version will take about six weeks.” That one sentence gives the speed people what they need — momentum and a real test — and gives the stability team what they need — time and a real process. Both loops, acknowledged. Nobody has to fight.
Key Takeaways
- Most AI friction inside accounting firms isn’t about AI — it’s a communication failure between two groups optimizing for different things.
- The two-loop model: Loop 1 fights uncertainty with speed; Loop 2 fights complexity with stability. Both are rational and necessary.
- If you’re in a stability role, translate your concerns into Loop 1 language. “What’s the minimum we need to test this?” moves conversations that “this adds complexity” never will.
- If you’re in a leadership role pushing AI adoption, explicitly decouple the loops. Give your stability team a real timeline — not just pressure to keep up.
- The most valuable AI skill in professional services right now is organizational translation: bridging the two loops so your firm can move fast and stay stable at the same time.
Want the CPE credit? Take the full lesson on EverydayCPE and earn 0.2 CPE credits: [lesson link]
Leave a Reply