Why AI’s next big opportunity lies in the technology businesses already use
Tech Funding News Andrea Mostosi ● Covered by 4 sources
AI is easy to try, but hard to wire into real work. Circeus says the next win is inside the software businesses already use, not another chatbot.
Based on reporting by Tech Funding News, Andrea Mostosi — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
A lot of companies can now use AI to draft an email. Far fewer can let it change how an order gets handled, a customer issue gets resolved, or a supplier gets paid. That gap is the whole bet behind Circeus’s latest case for business AI.
In a new piece by Andrea Mostosi, the London-based group’s head of AI argues that the problem is not simply model quality or reluctance from business owners. The real issue is where AI lives. If it sits in a separate chat window, someone still has to search for the records, paste them in, check the output and push the result back into the company’s systems. That is extra work dressed up as progress.
The adoption numbers in the piece explain why this matters. Official figures cited by Mostosi put AI use at between 17% and 20% of US businesses, and just under 20% of EU enterprises. The gap gets wider with size: 37% of US firms with at least 250 employees use AI, compared with fewer than 20% of firms with four or fewer. In the UK, around one in ten AI-using businesses with at least 10 employees described that use as extensive. Circeus’s own reading is harsher still: meaningful workflow implementation in the everyday economy is only about 4% to 5%.
Mostosi does not argue that model progress has stalled. He points to METR’s work showing task lengths AI can complete at a 50% success rate have expanded from minutes of expert work to hours on software and technical benchmarks. But benchmark success is not the same as safe delegation. For a 40-person distributor with no data team and no integration budget, the model is only part of the job. The rest is connecting it to the software where the work already happens.
That is why Circeus keeps coming back to accounting packages, booking systems and order-management software. Those tools already contain data, permissions and workflow logic. They also come with something a standalone AI product has to win from scratch: years of trust. Circeus says its own stack has three layers — the software knows, the model does, and the person decides — with evaluation built in by tracking task completion, error rates, human intervention and cost. If it cannot be measured, it does not ship.
The bigger twist is the ownership model. Circeus buys software businesses, then builds shared AI infrastructure across the portfolio instead of forcing each product team to reinvent the same safeguards and tooling. The pitch is simple: long-term ownership lets AI get embedded into the systems the everyday economy already runs on, rather than added as another project for customers to untangle.
My take — AI-written commentary, not fact-checked reporting
This is the part of AI that matters, and it’s the least glamorous one: boring software, deep integration, and a human still on the hook when things go weird. The chatbot era made everyone feel productive; the workflow era will expose who actually owns the process. Closed systems and acquired software may be a duller story than shiny model launches, but at least they have a shot at doing real work instead of demo theatre.
Read more about this at: Tech Funding News
