Speed without Clarity: Decision Debt in the AI Era

Written by Adrian Maharaj

(Views mine, not Google’s.)

I’ve been guilty of this. Not 5 years ago this year.
A colleague I trust pulled me aside: “You’re reopening the same decisions.”
They were right. I was socializing, perfecting, postponing paying interest on choices we’d already “made.” Naturally I started to look for patterns, both in myself and ways to broaden my own awareness.

This brought me to the numbers, and it wasn’t pretty. McKinsey estimates slow, inefficient decision making consumes ~530,000 manager days a year at a typical Fortune 500 about $250M in wages. That’s not a rounding error; that’s an operating model. (McKinsey & Company)

Zoom into a workday and the picture sharpens. Microsoft’s Work Trend Index shows the average employee spends 57% of time communicating (meetings, email, chat) and 43% creating; power users burn 8.8 hours/week on email and 7.5 hours/week in meetings. 62% say they lose time hunting for information. We’re talking about work more than we’re moving it. (Microsoft)

There’s a simple physics to why this drags. Little’s Law: more work‑in‑process → longer cycle times. Reopened or undecided calls keep work stuck in WIP; projects age, risk compounds, and teams start optimizing locally just to cope. I didn’t need a new framework to see it; I could feel it in my calendar. (Corporate Finance Institute)

Culture either compounds it or clears it. DORA’s long‑running research finds “generative” (high‑trust, learning) cultures correlate with ~30% higher organizational performance. Translation: healthier decision habits aren’t “soft stuff.” They show up on the scoreboard. (Dora)

Where this really bites is pilot purgatory. We trial, extend, re‑baseline then trial again. MIT Sloan’s research: only 16% of corporate ventures actually scale. That’s not always an idea problem; often it’s a decision problem no dated by call to scale, kill, or pivot. I’ve done this: kept experiments alive because deciding felt scarier than another “safe” month. (MIT Sloan Management Review)

How AI makes it worse (a scene from integrations)

AI makes integrations feel “instant,” which is exactly why decision debt hurts more.

A future not to far away…like months. Imagine a world where you tell a partner you’d adopt their new API version “soon.” Your doc bot rewrote the internal guide as if “soon” meant now. The test harness auto generated stubs and merged them to a feature branch. CS read the doc and promised a customer the integration was “live.” None of that was malicious. The system just treated an undecided decision as decided and sprinted.

By the time you corrected it, Support had tickets, Sales had promises, and the partner had expectations. No single step was the villain; speed without clarity was. What can you do to avoid this?

• Added an explicit decision state to every integration (proposed / approved / shipped)

• Tie automations to that state, and make “approved” impossible without a DRI (directly responsible individual) + date by. Less sexy than a new model.

• Way cheaper than apology emails. (FWIW: nearly half of workers now describe their work as chaotic and fragmented that feeling is exactly what happens when automation outruns alignment.) (Microsoft)

This is also why some big operators look “boring” from the outside and win anyway. PepsiCo’s AI programs are wired to measurable, financial impact before they scale demand forecasting, manufacturing, logistics, order management. It’s governance and workflow first, not tool FOMO. (The Wall Street Journal)

And the broad pattern is clear: the latest McKinsey AI survey ties workflow redesign and CEO‑level oversight of AI governance to higher self reported bottom line impact. The lever isn’t just what tool you pick; it’s how you wire decisions, ownership, and guardrails around it. (McKinsey & Company)

What I’m doing (and what’s started to work)

• Name the debt. For each big bet I ask: What’s the next irreducible decision? Who owns it? What’s the date by? If those answers wobble, it goes on a visible debt list. Right now I keep it close to the chest, but the goal by end of the year is to make it public

• Make decisions legible. A simple log: DRI, rationale, date, expiry, and kill/continue criteria. Not a fancy tool a table the whole team can see. WIP

• Classify the doors. Two‑way (reversible) → decide fast with a rollback plan. One‑way (irreversible) → slower by design. Objective: share this as a decision principal to my leadership team

• Measure the interest. Track decision latency (signal → decision) and reopen rate (calls revisited in 60–90 days). When both dropped, work started to flow. The mood changed, too.

• End pilot drift. Every pilot carries a scheduled scale/kill/pivot call before kickoff. Tie automations to the decision state so bots don’t outrun the humans.

Two predictions I’ll stand behind

1) Decision Latency will be a board metric by 2026. You’ll see “time to change” (signal → shipped change) next to NPS and CAC. The first firms to cut it by ~50% will show persistent AI attributable margin lift because they redesigned the work, not just the dashboard. Basis: McKinsey’s correlation between impact and workflow redesign + top house governance. (McKinsey & Company)

2) Reopen Rate becomes the quiet KPI for trust. Teams that stop reopening settled calls free up focus, reduce WIP, and shorten cycle time. It’s the simplest proxy for decision health I’ve found. Basis: Little’s Law + what you can feel in your calendar. (Corporate Finance Institute)

I’m not pretending I’ve solved it. I’ve just stopped letting it compound in the dark.
If this sounds familiar, try the three questions on your biggest bet: What’s the next irreducible call? Who owns it? When does it expire? If those answers aren’t crisp, that’s your decision debt. Pay it down before it compounds again.

Sources to keep you honest: McKinsey on decision cost (530k days / $250M), Microsoft Work Trend Index (57/43 and time‑searching), DORA (generative culture → ~30% higher org performance), MIT Sloan (only 16% of ventures scale), McKinsey 2025 State of AI (workflow redesign + CEO governance correlate with impact), Microsoft WTI on “chaotic work,” and queueing fundamentals via Little’s Law. (McKinsey & Company, Microsoft, Dora, MIT Sloan Management Review, Corporate Finance Institute)