← All articles

September 21, 2026

What to Write Down Before You Hire Someone to Automate Your Sales Workflow

Write a brief naming one workflow, the systems it touches, who owns it after go-live, and what done looks like, before you post the job.

ST

Stuart

Founder, Hotkey

Two job posts, same week, same job. One says “AI automation specialist needed, $25/hr, ongoing, must know AI and CRMs.” The other says “when a form is submitted on our site, enrich the lead, write it to the CRM, and draft a first reply for a human to approve within 10 minutes.”

The first one collects a crowd. The second one collects a short list you can actually read.

Write the brief before you write the job post. Seven lines: the one workflow, what triggers it, the systems it touches, what done looks like, who approves anything, the monthly running cost, and who owns it the Monday after go-live. These projects fail on scope and ownership far more often than they fail on the builder’s skill, and the brief is where both get decided.

Dozens of applicants, and nothing to sort them on.

We read a lot of automation job posts, on the public marketplaces where small firms hire for this work. The pattern doesn’t change: a post written as a job title pulls a large, undifferentiated field, and a post written as a workflow pulls a small one where every reply answers the same question.

The reason is mechanical. When the post says “automation specialist, hourly, ongoing,” each applicant has to guess what you want, so each one guesses differently. You end up comparing dozens of answers to dozens of different questions. There’s no shortlist to build because there’s nothing to sort on except price and profile polish, which are the two worst signals available.

Seven lines, and the one that does the most work.

An automation brief is a one-page description of a single workflow that states what starts it, what it touches, what finishes it, what it costs to run, and who owns it afterwards. Seven lines is usually enough. If you can write those seven lines, you’re ready to hire. If you can’t, the missing hour is yours, not a supplier’s. You can pay someone to help you write it, and plenty of good builders sell a fixed-fee scoping session, but buy that knowing what it is: an hour of someone else’s structure imposed on knowledge only you have. That hour is worth buying. Just don’t expect it to hand you a workflow you hadn’t noticed you had.

The trigger line does the most work of the seven. A lot of what gets described as an automation problem is really a question about when something should happen, and a workflow that fires on a form submission behaves very differently from one that fires at 8am every Monday. That’s your call, not the builder’s, because it depends on how your team actually works.

If you’re stuck on which workflow to put in line one, the shortcut isn’t to pick the task that eats the most hours. We wrote about choosing the first workflow separately, and the short version is that the best first candidate is the one that fails most visibly, not the one that takes longest.

How do you tell a good automation builder from a bad one?

Once the brief exists, grading candidates gets much easier, because you’re no longer assessing whether someone seems competent in general. You’re checking whether their answer to your seven lines is better than the other answers to your seven lines.

Four things separate the good replies.

They ask about your data before they talk about tools. Dirty records are the most boring reason these builds disappoint and one of the most common, and a builder who’s been burned by it asks early.

They name the failure mode without being prompted. Ask what breaks this workflow and a good answer is immediate and specific: the enrichment provider returns nothing for a third of records, the form sends a duplicate, someone renames a field in the CRM and the whole thing silently stops. An answer that describes only the happy path is a warning.

They push back on part of your brief. A brief written by a non-technical founder usually contains one requirement that’s disproportionately expensive for what it delivers. Someone who spots it and says so is doing the job. Someone who agrees to everything is quoting, not thinking.

They separate build cost from run cost without you asking. A builder who quotes only the build has either never run one of these in production or is counting on you not asking in month two.

How much does it cost to automate one sales workflow?

Nobody can tell you what automating one sales workflow costs, and the people publishing ranges are mostly publishing their own price list. Layer3Labs, an AI consultancy, publishes 2026 bands of $15,000 to $75,000 for a single workflow implementation, and $5,000 to $20,000 for strategy only if all you want is the plan. Those numbers are real, and they’re shaped for a buyer with a procurement process and an internal IT team. A 25-person service firm reading that page either overpays or concludes the whole category is out of reach, and both of those outcomes are wrong.

We’re not going to print a market rate we haven’t measured. What we can tell you is what moves the number. How many systems the workflow crosses. Whether the data in them is clean. Whether a human approves each run or it goes unattended, because unattended work needs error handling, retries and alerting that supervised work doesn’t. And whether you want it documented well enough that someone else could pick it up.

Whatever the workflow runs on has its own monthly bill, and it’s yours, not the builder’s. As of September 2026, n8n lists Starter at 20 euros a month billed annually for 2,500 workflow executions, and Pro at 50 euros a month billed annually for 10,000, with unlimited active workflows on both. Zapier prices on tasks rather than executions, and a task is only burned when an app action succeeds. Triggers are free, and so are built-in steps like filters, paths and Formatter. A Zap whose trigger fires seven app actions burns seven tasks where an execution-based plan counts one. The two models aren’t comparable on headline price alone. Ask for the expected monthly run cost in writing, at the quoted volume, before you sign anything. A build price with no run cost attached is an incomplete quote.

Ownership is three things. The third is the one nobody writes down.

You own the automation after the contractor finishes, and that has to be settled before the build starts rather than at handover, by which point you’ve nothing left to trade. Ownership is three concrete things and not a sentiment. The accounts and API keys are registered to your company, not to a personal address belonging to the person who built it. The documentation exists somewhere your team can find without asking. And one named person inside your business understands what the workflow does well enough to say what’s gone wrong when it stops.

The published advice here is better than it is on cost. Layer3Labs, to their credit, states plainly that you should own the system, understand how it works, and be able to maintain it without the consultant on retainer, and warns against proprietary lock-in. That’s right. What it skips is the third item, and without that one the other two are worthless. Owning credentials to a system nobody understands isn’t ownership. It’s an inheritance.

This is the same failure we wrote about when the person who built your automations leaves, and it costs nothing to prevent at brief-writing time. Asking who holds the accounts is one line. Asking for it after go-live is a negotiation.

When you shouldn’t hire anyone at all.

Some of what gets posted as an automation project is a checkbox nobody went looking for.

Modern CRMs ship with workflow rules, lead assignment, form routing and email sequencing, usually above the entry plan. Calendar tools have native meeting scheduling. Form builders have native CRM integrations. Before you write a job post, spend an hour in the admin section of the tools you already pay for and check whether the thing you’re about to buy is sitting there switched off. Paying someone to rebuild a feature you own is a common and entirely avoidable waste, and no supplier is going to point it out for you.

Don’t hire when the volume is too low to be worth clearing either. If the workflow fires twelve times a month and a person clears the lot in half an hour, you’re buying back six hours a year, and no build fee survives that arithmetic. Under a couple of hours a month it rarely works at any price, and the workflows that feel constant often turn out to sit still once you count them.

Hire when the workflow crosses systems that don’t talk to each other, when it needs judgment a rule can’t express, or when it has to keep running without anyone watching it. Don’t hire when the answer is a checkbox, and don’t hire when you can’t yet name the workflow, because that case produces a build that technically works and nobody uses.

The brief is the deliverable. The hire is only worth making if the brief still looks worth paying for once it’s written down.

Frequently asked questions

What should I include in a brief for an automation project?

Seven lines: the one workflow, what triggers it, the systems it touches, what done looks like, whether a human approves anything, the monthly running cost, and who owns it after go-live. Write all seven before you post the job. A brief that names one workflow end to end gets you a handful of comparable replies, where a post asking for an automation specialist by the hour gets dozens that each answer a different question.

Who owns the automation after the contractor finishes?

You should, and it needs to be written down before the build rather than negotiated at handover. Ownership means three specific things: the accounts and API keys sit in your company’s name, the documentation lives somewhere your team can find it, and one named person inside your business knows what the workflow does when it breaks. A build you can’t maintain without the person who made it is a dependency, not an asset.

Should I hire a freelancer or an agency to automate our CRM?

For a single workflow at a 10 to 50 person firm, a freelancer on a fixed scope is usually the better fit, because an agency prices in account management you don’t need for one build. An agency earns its premium when the work spans several workflows and several systems, and needs someone available when the one person who built it is on holiday. Either way the deciding factor is the brief, not the supplier type.

What questions should I ask an automation consultant before hiring?

Ask what breaks this workflow, who gets told when it fails, what it costs to run per month on top of the build, and who holds the accounts afterwards. A good builder answers the failure question first. Someone who only describes the happy path hasn’t built many of these, or isn’t thinking about the version you have to live with.

Do I need a developer to automate our sales follow up?

Often not. Lead routing, form-to-record mapping and follow-up sequences are usually already in the CRM you pay for, sitting switched off. Spend an hour in the admin section of your existing tools before you write a job post. The feature you are about to pay for may already be switched off in a product you own.

— Stuart, Hotkey