In the space of one week, Meta launched Muse, OpenAI showed off Dots and Microsoft slipped out WorkIQ. Three of the biggest companies in tech, all asking us to build the next generation of software on their agent platform.
We've seen this film before. Cloud had one. Mobile had one. Each time, the vendors fought hard for developers, and each time the eventual winner ended up somewhere most people never think about.
I think agents will go the same way. The interesting question is what that means for the businesses buying them.
For the last couple of years the conversation has been about models. Which one tops the benchmarks this week. Which one writes better code. Which chatbot window you open in the morning.
That's fading fast. The talk has moved to agents, and I suspect that's a stepping stone too. Soon we'll stop talking about either and go back to talking about things that get stuff done. They'll just happen to use agents underneath.
Google worked this out early. Most people using Google are using Gemini and have no idea. Search, Android, the assistant, it all runs on the same thing, and Google rarely makes a song and dance about it. They sell the outcome.
Meta is borrowing that playbook. If you use WhatsApp or Facebook, or one of their new glasses, you'll be using Meta's agent framework without ever choosing it. You'll ask for a hotel and a train ticket, and something in the background will sort it out.
That's what winning looks like. The platform that disappears into the products people already use.
For consumers, the platform question doesn't matter. Netflix runs on AWS and nobody has ever cancelled a subscription over it. People want the film. With agents, they'll want the train ticket.
For anyone building technology, it matters a lot.
Every one of these platforms is designed to capture you. Your agents, your data, your workflows, your context, all living inside someone else's walls. We spent a decade learning that lesson with AWS and Azure. Multi-cloud became a whole discipline because companies woke up to how hard it was to leave.
Most organisations have built their agent capability the other way round so far. They've stitched together parts that make sense for them: a model here, a harness there, their own tools and rules. That's how we've built ours at Dootrix. It's flexible, and it's ours.
A platform war puts pressure on that. The convenience of an all-in-one agent service is real. So is the cost of handing over the thing that's quickly becoming your most valuable operational asset.
My advice to any CIO is simple. Own your harness. Keep your context and your rules portable. Treat the model as a component you can swap, because you will want to.
There's one thing in OpenAI's Dots announcement I think everyone should copy, whichever platform they pick.
Most personal agents today act as you. You hand over your logins and the agent goes off and pretends to be you. It reads your email as you and opens your documents as you.
Dots gives each agent its own account. The agent becomes a user in your organisation, and you grant it access the same way you'd grant access to a new member of staff.
That's a far better security model. You can lock down exactly what an agent can see and touch. You can audit what it did. You can switch it off without touching anyone else's access. Zero trust by default, with granular permissions, the way we'd treat any new hire.
If you're building systems with agents in them, they should already have their own identity. If they don't, fix that before you worry about which platform to back.
The part none of the launches answered is the money.
An always-on agent with its own compute, learning over time, is expensive to run. Plenty of people already struggle to justify twenty dollars a month for a chatbot. They aren't going to pay two hundred to have an agent check whether the coffee shop is open.
That suggests two possible paths. Either these platforms follow social media and the user becomes the product. Or they give it away until people are hooked on the productivity, then raise the price. Anyone who has built a proper second brain knows how hard that would be to give up.
For businesses, the maths is different. The value is easier to see and easier to measure. Which brings us back to lock-in. The more value your agents create, the more it matters who owns them.
The agent platform war is real, and it's only just starting. The winners will be the platforms that quietly power everything, and most consumers will never know their names.
Businesses don't get that luxury. Decide now what you need to own, give your agents their own identities, and keep your options open. The vendors will keep shipping. Your job is to make sure you can move when they do.