The Frontier Firm isn't a tool stack. It's a decision.

Microsoft has been pushing a phrase recently: the Frontier Firm. It is a useful label, and like most useful labels, it is in danger of becoming a marketing wrapper for nothing in particular.

The original idea is straightforward enough. A Frontier Firm is one where AI is woven into how people actually work. Not bolted onto the side, not stuck in a pilot, not living inside one team while everyone else carries on as before. The phrase points at a real distinction. There are firms where AI has changed the work and firms where AI has merely been deployed. The gap between those two states is enormous.

What gets lost in the marketing version is that you do not arrive at this state by buying enough things. You arrive at it by making a series of decisions, most of which have very little to do with technology.

What firms on the frontier have actually decided

When I look at the firms I would describe as genuinely operating on the frontier (not just the ones with Copilot licences and good slides), they share a small number of commitments. Not capabilities. Commitments.

They have decided that “we don’t know yet” is an acceptable answer. The firms that are stuck spend a long time waiting for somebody to publish the playbook. The frontier firms accept that there isn’t one, that there might never be one, and that the only way to find out what works is to try things and see. That is uncomfortable for organisations that pride themselves on certainty, which is most professional services firms.

They have decided that experimentation is part of the day job, not a side quest. This sounds small. It is not. Most firms talk about “innovation” while structurally punishing anyone who spends time on it instead of billable work. Frontier firms have made deliberate choices (protected time, recognised contribution, real career consequences) that say spending your Thursday afternoon working out whether Copilot can do part of your job faster is exactly what we want you to be doing.

They have decided to share the things that didn’t work. The firms that get good at AI fastest are the ones where someone can say in a meeting “I tried X and it was useless” without it being a career problem. That sounds obvious until you watch how rarely it actually happens in firms that are stuck.

They have decided who owns this. Not IT alone. Not the partners alone. Not a steering committee. Someone has been given the job of making this work and has the air cover to actually do it. That person is usually the difference between a firm that experiments and a firm that adopts.

What the tools have to do with it

Tools matter. This is not an argument that they don’t. They just sit downstream of the decisions, not upstream.

A firm that has made the decisions above will get extraordinary value out of a relatively modest tooling stack: Copilot, a few well-built Power Automate flows, a SharePoint that doesn’t fight you, maybe one or two purpose-built Copilot Studio agents for the work that comes up again and again. None of this is exotic. All of it is available to firms that have not made the decisions, and they get very little out of it.

The frontier is not where the technology is. The frontier is where the willingness is.

Where this leaves you

If you are leading a firm and you are wondering whether you are on the frontier, the answer is almost certainly no, and that is fine. Almost nobody is yet. The more useful question is: which of the decisions above have we actually made, on paper, with consequences attached? Not “we agree in principle.” Made.

Most firms find that the answer is none of them. Which is the first place to start.

That first step is usually not another tool. It is a clear Copilot adoption strategy and a few well-chosen Power Platform automations that prove the new way of working in public.