The mistake this screen prevents
Automating a messy process does not fix the mess. It runs the mess faster, at volume, often in front of customers. That is the classic version. The expensive version is paying for a tool or a build on a workflow that ten minutes of honest checking would have disqualified: the process only exists in one employee's head, or half the cases are exceptions, or nobody can say who checks the output. The screen exists so the disqualifying answer arrives before the invoice, not after.
What changed recently
Automation used to be something you went out and bought, and the buying process forced a pause. Now the software you already pay for ships AI assistants and agent features as standard: the tools running your sales pipeline, your support inbox, and your books can draft, route, and act with a toggle. That is genuinely useful, and it removes the pause where second thoughts used to happen. When automating is one click, the audit has to be deliberate, because nothing in the product will ask whether your process is ready.
Pick the right workflow to screen first
Do not start with the workflow that annoys you most. Start with one that is frequent, repetitive, internal, and cheap to get wrong: drafting follow-ups, summarizing notes, first-pass sorting of an inbox. Frequency is where the payoff lives, repetition is what AI handles well, and a small blast radius means a mistake teaches you something instead of costing you a customer. The dramatic, customer-facing automation can wait until the boring one has proven the team can run this.
The five checks that disqualify a workflow
One: the process is not written down. If it lives in someone's head, you would be automating a guess. Document it first; that work pays for itself even if you never automate. Two: exceptions outnumber the rule. AI handles the routine case; if most cases are special, there is no routine to automate. Three: mistakes reach customers before a person sees them. Until a human checkpoint exists, keep AI on internal drafts. Four: the workflow touches sensitive data and you have no rule about which tools may see it. Set the data rule first. Five: nobody owns the output. If no one is named as the person who reviews and answers for the result, the automation will drift until it fails in public. Any one of these is a stop sign, and all five are fixable.
If it passes, what you have earned
A passing screen does not mean buy something. It means the workflow has earned the deeper audit: mapping the steps, the data, the failure points, and what good output actually looks like, before a tool is chosen. Run the screen on your top three candidate workflows this week; it costs an afternoon. Do not act yet if every candidate fails, if the team is mid-turnover, or if the only reason is that the software turned the feature on. And if the workflow that passes is one where the stakes feel high or the path is unclear, that is exactly the moment a short session with someone who does this weekly saves you from learning on production.
The short version
- Automating a messy process makes the mess faster. Screen before you spend.
- AI features now ship inside software you already pay for, so the pause before automating has to be deliberate.
- Screen the frequent, repetitive, internal workflow first, not the dramatic one.
- Five disqualifiers: undocumented process, exceptions outnumber the rule, mistakes reach customers, sensitive data without rules, nobody owns the output.
- Passing the screen earns the deeper audit, not the purchase.