The cost of waiting for clarity

Let me start with a question.

What has your organisation actually done about AI in the last six months?

Not planned. Not discussed at a board meeting. Not added to a strategy document. Actually done.

Hold that answer. We’re coming back to it.

I’ve been watching AI closely for years. I write about it over on my personal Substack Antonym, build with it, and advise organisations on it. And I’ve learned to distinguish signal from noise – because there’s a lot of noise.

So when I tell you that something genuinely changed in the last eight weeks, I’m not saying it to get your attention. I’m saying it because I think most leadership teams haven’t had a chance to actually take stock of it, and the cost of not doing so is now measurable. I think this moment is very similar to when ChatGPT came out.

Here’s what happened.

In early 2026, enterprise software companies lost roughly $2 trillion in market value. Salesforce is down over 30%. Atlassian, Intuit, ServiceNow – all hit multi-year lows in a matter of weeks. Investors are pricing in a world where AI agents replace per-seat subscriptions.

Think about that for a moment. The software your organisation spends the most on, per head – investors are now betting it faces its first real competition in thirty-five years.

That isn’t analysts speculating about the future. That’s markets making a call about this year.

In January, Anthropic launched Cowork with plugin support. Within six weeks, it moved from a chat tool to an operational platform – connecting AI agents to Slack, Asana, Google Drive, and dozens of others. Then enterprise plugins launched for finance, HR, design, and operations. The progression from “AI chatbot” to “agentic platform with an ecosystem” happened in less time than most organisations take to approve a new software trial.

This is what I’d call the app store moment for AI. Skills and plugins mean the distance between “we want to do X with AI” and “we’re doing X with AI” is now measured in days, not months. And you don’t need to write code to get there.

Then there’s OpenClaw. An open-source autonomous AI agent built by one developer in Austria. It became the fastest-growing project in GitHub history – 190,000 stars in 14 days, overtaking React, a project developers have been voting for since 2013. In its first week it also exposed 1.5 million database records and enabled an $8 million crypto scam. Sam Altman hired the creator to lead OpenAI’s next-generation personal agents programme.

The future of AI isn’t tidy. But you can see it taking shape.

Here’s the framing I keep coming back to. There’s a scene in Interstellar – Miller’s world – where an hour on the planet surface is seven years back on the ship. That’s roughly what’s happening with AI adoption right now.

One team spends a quarter deliberating about governance frameworks. Another runs a thirty-day experiment, generates real data, and makes a decision. The gap between them isn’t linear. It’s gravitational.

McKinsey reports that only 23% of enterprises are actually scaling AI agents. 65% cite expertise gaps.

The gap between what’s technically possible and what most organisations are doing has never been wider. And it isn’t a technology problem. The technology is here. It’s accessible. You don’t need a machine learning team or a transformation programme to start.

It’s a decision problem.

The organisations I worry about aren’t the ones that have tried AI and found it wanting. They’re the ones that are waiting for clarity before they act. Waiting for the technology to settle. Waiting for a strategy that applies to everyone. Waiting for certainty that isn’t coming.

Because here’s the thing about the current moment: the clarity you’re waiting for will only come from doing. Not from watching, not from reading briefings, not from convening a working group. From running real experiments with real work and seeing what happens.

I’m not going to give you a twenty-point plan. If you leave this post with three things you actually do, it will have been worth your time.

Today. One hour. Take a task you’d normally spend an hour on – a briefing note, a summary, a first draft of something. Run it through Claude or ChatGPT. Don’t give it a test task. Give it something real and see what you get back. That experience – whatever happens – is your personal baseline for what’s possible right now.

This week. Two hours with your team. Run an honest session with one question: where are we actually with AI? Not where the strategy says we should be. What are people already using? What’s blocked? What would they try if someone said yes? Almost every organisation I work with has shadow AI use – people using personal accounts to do work the organisation hasn’t sanctioned. That’s a signal, not a problem. Surface it.

This quarter. One experiment. Go back to the software question – what are you paying most for, per head. Pick one workflow inside that tool. Find an AI agent alternative. Run a thirty-day parallel trial with a willing team. Measure time, cost, quality. That gives you real data for a bigger decision. Not a pilot programme. Not a strategy. One workflow, one team, thirty days.

The organisations that will make good decisions about AI over the next two years are the ones generating real data from real experiments. Not the ones with the best strategy documents.

On Wednesday 11 March at 11am GMT, I’m running a free 30-minute session called ‘AI Reality Check: March 2026’.

I’ll cover what’s actually changed and why it matters, what agents, skills, and plugins mean in plain English, what the market repricing signals for your technology spend, and the three actions above in more detail – with time for your specific questions.

It’s 30 minutes of peer-to-peer conversation for senior leaders who want to get clear, not just get current.

If you can’t attend live, register anyway – I’ll send the recording within 24 hours.

Register here

And if you want something tailored – mapping this directly to your sector, your tech stack, your team – I have a handful of 90-minute private briefing slots available this month. Email hello@brilliantnoise.com.

Antony Mayfield is the founder of Brilliant Noise. We’re an AI partner and consultancy that helps organisations build lasting AI capability.