--- title: "We Build Amy to Fit the Job, Not the Other Way Round" description: "Professionals are adopting AI faster than their organisations can decide on it. Why we start with the person doing the work rather than the person signing the contract." category: "Posts" date: "2026-08-17" author: "Gabriel" image: "/blog/we-build-amy-to-fit-the-job.jpg" tags: ["AI adoption", "product", "future of work"] --- A customer signed up for Amy last week, connected their own tools, set up routines around their own week, and moved to a paid plan two days later. They didn't speak to anyone from our team. The speed is not the interesting part. The interesting part is that nobody had to explain to them where Amy belonged in their job, and nobody had to ask them to work differently to get the benefit. They described the work that they're already doing, in their own words, and delegated it to Amy. That is who Amy is built for, busy professionals, each with a different job, none of whom should have to reshape that job around the product. ## Two people, two different questions Ask a head of operations what their team needs from AI and you get a good answer. Where capacity is short. Which work is inconsistent between people. What the team never gets to. Where the risk sits. They know their function, and that answer is the right input for deciding what to prioritise. Ask the analyst what they need from AI and you may get a different answer. Spending 3 hours every Monday pulling the same numbers from four different tools into a standardized document. Responding to different stakeholders with project updates. The update takes forty minutes, and 35 minutes for them to assemble. Both are reasonable responses. They are answers to the same question with different core responsibilities. The person best placed to describe that task is precisely the person who does the required work every week. So the two needs are not in tension. They are in sequence. The outcome the manager wants is produced by tasks changing hands, and the people holding those tasks are the ones who can specify them. ## The needs differ by role, so the product has to This is the part most B2B software gets wrong in the world of AI. The traditional model is that a company buys software, and then the roles adapt to it: new fields, agreed conventions, a way of working that exists because the software needs it to. Sometimes that is worth building out a process. It is also why so many AI rollouts have failed. AI doesn't work like that, because the unit is the task, and tasks are role-specific, where software is not. The analyst's Monday numbers, the associate's diligence summaries, the ops lead's onboarding checks, the founder's investor update: same team, same week, four completely different jobs. A single prescribed workflow serves none of them properly. The test we hold ourselves to is simple. If somebody has to change how they work in order for Amy to be useful, that is our design problem to fix, not their adoption problem to overcome. ## The workforce wants to leverage AI, and the data says so Professionals have already worked out that these tools are useful, and they have not waited for the companies to catch up. Microsoft and LinkedIn found that 78% of people using AI at work were bringing their own tools to it, rising to 85% of Gen Z ([Work Trend Index, May 2024](https://www.microsoft.com/en-us/worklab/work-trend-index/ai-at-work-is-here-now-comes-the-hard-part)). Menlo Ventures traced the money to the same place: 27% of enterprise AI application spend now arrives through product-led channels, against roughly 7% for traditional software, and they note that once you count personal cards the real figure is probably closer to 40% ([Menlo Ventures, December 2025](https://menlovc.com/perspective/2025-the-state-of-generative-ai-in-the-enterprise/)). That is a workforce running its own experiments, on its own time, on the work it knows best. The organisation around them is usually still strategizing. Microsoft's 2026 Work Trend Index, across 20,000 workers in ten countries, found only 19% of AI users sit in what they call the "frontier zone", where individual capability is matched by real organisational support. Just 26% say their leadership is clearly and consistently aligned on AI ([Microsoft, Work Trend Index 2026](https://www.microsoft.com/en-us/worklab/work-trend-index/agents-human-agency-and-the-opportunity-for-every-organization)). That is not a criticism of anyone's leadership. Getting an organisation aligned on a technology moving this quickly is genuinely hard with multiple stakeholders with different wants and needs. It does mean the person with the clearest view of where AI helps is often waiting on a decision. ## Most of the market is still building for the buyer Almost everyone selling AI to businesses is still running the B2B company-first playbook: seat licences, committee pilots, rollouts designed to be approved. Six to ten decision makers shape a typical purchase, each arriving with their own research ([Gartner](https://www.gartner.com/en/sales/insights/b2b-buying-journey)). Oversight is a real need, and a manager asking for visibility is asking for something reasonable. The failure mode is when a product is built for that need alone, because then it optimises for the person who signs rather than the person who uses, and nothing changes Monday for the team. Millennials and Gen Z now make up 71% of B2B buyers, and nearly half of business purchases are already self-service ([Forrester, March 2024](https://www.forrester.com/blogs/younger-b2b-buyers/)). This is a cohort accustomed to researching and trying new tools for themselves. ## Why we start with the person doing the work Three things have to be true for AI to change how work actually gets done. Someone has to know the work specifically, not categorically. Someone has to want the outcome. Nobody needs a change programme to persuade them to stop doing the part of their job they resent. Someone has to be able to check the quality of output. This is the one most people underrate. Delegation only saves time if verifying the result is faster than doing the work, and verifying requires the expertise to know what right looks like. Budget and authority can be supplied from above. Those three cannot. They sit with the professional doing the job, which is why that is where we start. ## This is not an argument against selling to companies Teams buy, and we help people directly the whole way through: any user can book a session with an Amy Expert whenever they want, because the "ok, what else can I do?" moment can be a barrier and speaking to someone is still the fastest way through it. What changes is the order. The manager's outcome and the analyst's task are the same problem seen from two ends, and the case for the company gets made by somebody inside it who has been using the product every day. We earn the right to sell to the company, one user at a time.