Most AI automation projects don't fail because the technology isn't good enough. They fail because the brief was vague. "We want to automate our workflows" tells an agency almost nothing. It's the equivalent of walking into an architect's office and saying "we want a building." The more precisely you can describe the problem you're trying to solve, the faster an agency can build something that actually works — and the less you'll spend going back and forth on revisions. Here's exactly what to include in your brief to get the best results from day one.
Start With the Problem, Not the Solution
The single biggest mistake clients make is leading with the tool they think they need. "We want a chatbot" or "we want something in Make.com" jumps straight to the solution before the problem is properly defined. Start instead with a plain-English description of the pain.
Think about it this way: what's the task that's quietly eating your team's time every week? For a lot of businesses, it's something mundane — manually copying data from one system to another, chasing clients for documents, sending follow-up emails that everyone knows need to go out but keep falling through the cracks. One London-based property management firm was spending 11 hours a week just reconciling maintenance requests from email, WhatsApp, and an online portal into a single spreadsheet. They didn't know what automation they needed. They just knew it was painful.
In your brief, describe:
- What the task is, step by step, as if you were explaining it to a new employee
- How often it happens (daily, weekly, per new client, etc.)
- Who currently does it and roughly how long it takes them
- What goes wrong when it's done manually — missed steps, delays, errors, complaints
That last point matters more than people think. Agencies can build faster and more robustly when they know the failure modes upfront. If the process breaks every time a client submits a form with a missing field, say that.
Map Your Existing Tools and Where the Handoffs Break Down
AI automation doesn't replace your existing software stack — it connects it. So an agency needs to know exactly what tools you're already using and, critically, where the handoffs between them are manual.
List every piece of software involved in the workflow you want to automate. This doesn't need to be technical. Something like: "A client fills out a Typeform → we copy the details into HubSpot → someone emails a welcome pack from Gmail → the project gets added manually to Asana" is a perfectly useful workflow map. You've just shown four systems and three manual handoffs, each of which is a delay and a potential error point.
Be specific about versions and plans where you can. Some tools only expose their data to automation platforms on higher-tier subscriptions — your agency will want to check this early rather than halfway through a build.
Also flag any tools that are off-limits or awkward. A medical clinic, for example, might use patient management software that has strict API restrictions for compliance reasons. A law firm might have a document management system that IT won't allow third-party connections to. Surface these constraints early and you'll save everyone a lot of time.
Define What "Done" Looks Like
Agencies need a clear definition of success — and so do you, frankly, because it's the only way to know whether the automation is actually delivering value after it goes live.
Be as specific as possible. "We want this to be faster" is not a success criterion. "We want the time between a new lead filling out our contact form and receiving a personalised response email to drop from 4 hours to under 5 minutes" is. "We want our team to stop manually entering data into the CRM" is better, but "we want zero manual data entry for new client onboarding, with all fields populated automatically from the intake form" is better still.
If you have data, use it. Here's why this matters: one e-commerce brand that briefed BrightBots knew they were losing roughly £8,000 a month in abandoned orders because follow-up emails were being sent too slowly — sometimes days after the cart was abandoned. Because they could quantify the problem, we could build toward a clear ROI target. The automation we built cut follow-up time to under 15 minutes and recovered around 22% of those carts within the first two months. That's the kind of outcome that's only possible when the brief includes real numbers.
Think about:
- Volume: how many times per day/week/month will this automation run?
- Speed: how fast does each step need to happen?
- Accuracy: what error rate is acceptable, and what happens when something goes wrong?
- Notifications: does a human need to be alerted for exceptions, approvals, or edge cases?
Be Honest About Budget, Timeline, and Internal Capacity
This section makes clients uncomfortable, but skipping it causes more problems than it solves.
You don't need a precise budget to the penny, but giving a range helps an agency scope the right solution. A £2,000 budget and a £20,000 budget will produce very different builds — and neither is wrong, they're just different. A smaller budget might mean a leaner automation that handles the core 80% of cases; a larger one might mean full exception handling, a dashboard, and ongoing monitoring. An honest conversation about this upfront stops you getting a proposal you can't afford or, worse, a stripped-back build that doesn't actually solve the problem.
Timeline matters too. Automations built under artificial pressure tend to have gaps. If you need something live in two weeks because of a product launch, say so — but also understand that might affect what's possible. Most well-scoped automations take between two and six weeks to build and test properly, depending on complexity.
Finally, be realistic about what your team can handle. Some automations require a human in the loop for approvals or exceptions. If the person who would handle exceptions is already overloaded, that's worth flagging. The best automation in the world fails if nobody is available to manage the edge cases.
Conclusion
A well-written brief is genuinely the highest-leverage thing you can do before engaging an agency. It compresses timelines, sharpens the build, and means the final product actually solves the right problem. You don't need to know anything about AI, APIs, or automation platforms to write a good one — you just need to know your own processes, your tools, and what success looks like. Spend an hour documenting those things clearly before your first conversation with an agency, and you'll get better results, faster, for less money.