Here's an uncomfortable truth about AI in mid-2026: if a process touches your business, it can probably be automated. The technology stopped being the bottleneck a while ago. Which means "what can we automate?" no longer separates good decisions from bad ones — everything qualifies.
The question that actually matters is: what should you automate? And just as important — what shouldn't you touch at all?
That's why every process we examine in an audit receives exactly one of four verdicts. Not a score, not a maturity rating, not a slide of possibilities. A verdict. Here's how each one works, and how to tell which one a process deserves.
Verdict 1: Automate
AI or automation runs the process end to end. A human reviews exceptions only.
A process earns this verdict when it checks the boxes: repetitive, rule-based, digital inputs and outputs, low judgment, and — critically — errors are recoverable. Invoice reconciliation. Follow-up sequences. Status updates. Report generation. Work that never should have consumed human hours in the first place.
Notice what the verdict requires. It's not "AI could do this." It's "this process is defined tightly enough, and the cost of a mistake is low enough, that AI should own it." If you can't describe the rules, the process isn't ready for this verdict — it's ready for documentation first.
Verdict 2: Assist
AI prepares. A named person decides.
This is the default verdict for judgment work with mechanical prep. The proposal that takes three hours to draft but ten minutes to approve. The meeting summary that eats an afternoon but informs a two-minute decision. AI eliminates the drafting, the summarizing, the research — the slow part — and hands a decision-ready package to a human.
The rule that makes this verdict real: a name, not "the team." "AI drafts, the team reviews" is how quality quietly dies — everyone assumes someone else read it. "AI drafts; Sofia reviews and sends" is a working system with accountability built in. If you can't name the person who stays in the loop, you don't have an Assist process. You have an Automate process pretending to be safer than it is.
Verdict 3: Human
The process stays fully human — deliberately, and permanently.
Not because AI can't do it. Because it shouldn't. High-stakes judgment. Ethics. Regulatory exposure. And above all: the moments where the relationship is the product. The difficult conversation with a long-time client. The pricing negotiation. The call where someone needs to hear a human voice deciding, not a system responding.
A Human verdict isn't a gap in your AI strategy. It's a strength you've named. When you can tell your team — and your customers — exactly which parts of your business will always have a person behind them, you've built something your competitors' chatbots can't copy.
Verdict 4: Leave Alone
Don't touch it. Not even to optimize it.
This is the verdict nobody selling AI will ever give you, which is exactly why it belongs in the framework. Some processes work fine. Some run so rarely that automating them costs more than a decade of doing them by hand. Some are held together by a person who's brilliant at them and would be demoralized by the change — and the disruption would cost more than the efficiency gains.
The cost of change is real. The best AI roadmaps account for it — and the honest ones include a list of things not to do.
One line of "why" is enough: works fine, low volume, not worth it. Then move on to the processes where the verdict actually changes something.
The Trust Test
Here's how to evaluate anyone advising you on AI — including us. Look at their recommendations and count the verdicts.
If every single process came back "automate," you're not being advised. You're being sold. A business where literally everything should be automated doesn't exist — and an advisor who found one wasn't looking, because every automate verdict happens to be billable.
That's why we hold ourselves to a standing rule: every audit names at least one thing to keep human or leave alone. If we can't find one, we look harder. The verdicts only mean something if all four of them get used.
Try It Monday Morning
You don't need us to start. Pick five processes in your business — the ones that eat the most time or cause the most friction. Give each one a verdict. Then ask the person who actually runs the process to do the same, separately, and compare.
The disagreements are the value. Where you said Automate and they said Human, there's a conversation about what that process actually involves — and that conversation usually reveals the business doesn't work the way the org chart says it does. Which is the real starting point for everything.
If you listed your ten most important processes right now — could you defend a verdict on each one? Not what AI could do. What your business should do.
Want a Verdict on Every Process You Run?
That's the BRP AI Audit: your business captured as it actually works, four verdicts applied, quick wins ranked, and the financial impact in dollars.
See How the Audit WorksOr if you'd prefer to talk it through directly, book a clarity call and we'll put verdicts on your five biggest time-eaters together.