Which processes to automate first? Rank candidates by value, effort, and readiness, then start with the highest-return workflow. A practical order.
A 25-person firm has more processes worth automating than it can tackle at once. Payroll runs late, invoices go out on memory, new clients wait on a manual onboarding checklist, and every month someone rebuilds the same report by hand. The hard part is rarely finding something to automate. It is deciding which processes to automate first, because the order you pick decides whether the project pays for itself or quietly stalls.
Quick Answer: Automate the process that is repetitive, rule-based, high-volume, and already digital, and that you can measure before and after. In practice that usually means billing follow-up, client intake, or recurring reporting. Rank your candidates by value, effort, and readiness, then start with the one highest-value, low-effort process and prove it before expanding.
Start with one process that repeats often, follows clear rules, runs on data already in your systems, and carries a cost you can measure in hours or errors. For most small firms that is billing follow-up, client intake, or recurring reporting. Pick the single process wasting the most predictable time, automate its handoffs, and prove the result before you touch anything else.
The reason to go narrow is that the opportunity is bigger than your capacity to act on it. McKinsey's November 2025 analysis found that demonstrated technology could already automate activities accounting for about 57 percent of US work hours. That is not a to-do list. It is a warning that you can scatter yourself across a dozen half-built automations and finish none of them. Starting with one process turns a vague percentage into a specific hour count you can recover.
The hours are easy to find once you look for them. In Smartsheet's workplace automation survey, more than 40 percent of workers said they spend at least a quarter of their week on manual, repetitive tasks, led by email, data collection, and data entry. Your first target is sitting inside that quarter of the week. If you already know the workflow and want it built, that is what our implementation work covers, but the ordering below comes first.
Score each candidate process on three things: value, effort, and readiness. Value is the hours saved plus the errors or risk removed. Effort is how many systems must connect and how complex the logic is. Readiness is whether the process is stable and you can write its steps down. The best first project scores high on value, low on effort, and high on readiness.
Run each candidate through this and the shortlist writes itself:
| Test | The question to ask | What a strong answer looks like |
|---|---|---|
| Value | How many hours or errors does this cost every month? | A number you can baseline now, not a guess |
| Effort | How many systems have to talk to each other? | Two or three, not your entire toolset |
| Readiness | Can you describe the steps to a new hire in writing? | Yes, and they hold true every time |
The trap most owners fall into is scoring on gut instead of numbers. The process that annoys you most is not always the one costing you the most. Spend the time to baseline two or three candidates properly. A firm that spends two weeks choosing the right process and eight weeks building it beats one that spends ten weeks building whatever looked impressive in the last software demo. For a fuller catalogue of candidates, see our guide to what to automate in your business.
The best first automations live in the handoffs between systems, where a person copies information from one place to another. In rough order of how often they pay off for small firms: billing and invoicing, client or lead intake, onboarding, and recurring reporting. Each is high-volume, rule-based, and already sitting in your software.
Here is why they tend to rank in that order:
In the law-firm audits we run, every one has uncovered ten or more automation opportunities. The pattern is that the highest-return ones are rarely the workflows the partners named walking in. That is the argument for scoring candidates rather than trusting the first idea.
Fix the data before you automate the workflow. If the same information lives in several systems that disagree, automating a step on top of that just moves bad data faster. Get one reliable source of truth first, then automate the highest-value handoff, prove it, and use that win to fund the next one.
A workable sequence for the first three months looks like this:
We built a single source-of-truth data layer for a 500-employee New York real estate company for exactly this reason. The first real problem was not a clever feature. It was that the same records lived in five places and never matched, so anything automated on top of them would have scaled the disagreement. Before your first build, it is worth checking whether your business is ready for AI at all.
Three failure modes show up again and again: automating a broken process so it breaks faster, chasing the exciting idea instead of the profitable one, and building on messy data. Each one turns a project that should have paid for itself into one that gets quietly shelved.
The abandonment is common enough to have a number. S&P Global found that the share of companies scrapping most of their AI initiatives climbed to 42 percent in 2025, up from 17 percent the year before. Reporting on that survey pointed to a clear split: the firms that avoided the scrap heap were the ones that prioritized their use cases rather than chasing every idea. Picking the wrong process first is how a promising project becomes part of that 42 percent.
Two honest "not yet" cases are worth naming:
Leave three kinds of process for later: anything that needs human judgment on every case, anything that happens too rarely to repay the setup, and anything still visibly broken. These look automatable in a demo and turn into maintenance work in real life. Park them, and start somewhere with clear rules and steady volume.
Quick filter for what to skip on the first pass:
Start where the rules are clear. The exciting work gets easier once the boring work already runs on its own.
Not sure which of your processes to automate first? That is the exact question our AI assessment answers. Six questions, a couple of minutes, and you get a personalized preview of where automation would pay off in your business, before you spend anything on an audit or a build. Start assessment. You can also browse more practical guides on the Eleventh AI blog, including what an AI automation audit includes when you are ready to go deeper than a single process.