Your AI Strategy Lives in Your Team's Backlog of Manual Work
Most AI strategies fail because they start in the wrong place. They start with a vendor demo, a McKinsey chart, or a competitor's press release. Then someone hands you a roadmap built on what AI could do, and six months later you have three half-finished pilots and a Notion page nobody reads.
Here is a better starting point: the backlog of manual work your team already hates doing. That backlog is the most accurate, highest-ROI AI strategy document you will ever produce. You just have not written it down yet.
Why the backlog beats the trend report
A vendor brief tells you what is possible. Your backlog tells you what is valuable. Those are not the same thing.
When your operations lead spends three hours every Friday reconciling invoices against a spreadsheet, that pain is real, measurable, and recurring. When your sales rep retypes the same client information into four systems, that pain has a dollar figure attached. When your bookkeeper exports a CSV, cleans it in Excel, and pastes it into QuickBooks, you are watching money walk out the door in fifteen-minute increments.
None of that shows up in a trend report. All of it shows up the moment you ask your team what they wish would go away.
The backlog also filters out the seductive stuff. Generative AI is good at a lot of things, but most small businesses do not actually need a customer-facing chatbot or an AI-written blog farm. They need their intake forms to stop dying in someone's inbox. They need their quote process to take twenty minutes instead of two hours. Those are not glamorous problems. They are the right problems.
How to actually capture the backlog
There are three ways to do this, and you should probably do all three. Each one surfaces a different kind of pain.
Run a structured session. Get the team in a room, or a Zoom, for ninety minutes. Ask one question: "What did you do this week that a reasonably smart intern could have done?" Then ask it again, phrased differently: "What task did you put off because it was tedious?" And again: "What do you do the same way every time?" Write everything down. Do not edit. Do not evaluate. The goal is volume, not quality control.
Send an async form. Not everyone thinks well in groups. Some of your best candidates will come from the quiet person who has been silently hating a process for two years. Send a short form, five questions, and give people a week. Ask for specific tasks, rough time spent per week, and the systems involved. The form catches what the meeting misses.
Sit with the team for a day. This is the one nobody does, and it is the most valuable. Pick the role with the highest manual load (usually ops, finance, or customer service) and shadow them for a day. You will see things they forgot to mention because the pain has become invisible to them. The triple data entry. The Slack message that triggers a manual export. The morning ritual of checking six dashboards because no one trusts the summary email.
By the end of this, you should have somewhere between thirty and a hundred candidate items. That is the raw material.
Score the backlog before you build anything
A list of complaints is not a roadmap. You need to rank. The framework we use at Catalyst on every engagement comes down to four variables, and you can run it in a spreadsheet in an afternoon.
Multiply time by frequency to get annual hours saved. Then filter for low variability and low failure cost. The items that survive are your first three workflows. Not your first ten. Three.
We wrote more about how to pick the first one in our piece on choosing your first automation, but the short version is: pick the boring one with the highest hours-saved score and the lowest blast radius if it breaks.
The counterargument, and why it is mostly wrong
The pushback on this approach usually sounds like: "If we only automate what we already do, we will miss the bigger opportunity. AI is supposed to change how we work, not just make the current way faster."
That is true in theory. In practice, almost no small business has the capacity to redesign their operations from scratch while also running the operations. You will not invent a new sales motion by reading Anthropic's product blog. You will invent it by automating the tedious half of your current sales motion, freeing up twelve hours a week, and using that time to think.
Transformation follows automation. It does not precede it. The companies that successfully reinvented their workflows with AI in 2024 and 2025 almost all started by quietly removing manual work first. That bought them time, credibility with their team, and a clearer view of where the actual constraints were.
There is also a cultural cost to starting with the visionary roadmap. Your team has been complaining about manual work for years. If your AI strategy ignores their list and arrives with a new one from a consultant deck, you have just told them their pain does not matter. Good luck getting adoption on whatever comes next.
What this looks like in practice
A recent client of ours came in wanting an AI agent to qualify leads. Reasonable ask. We ran the backlog exercise first. Twenty minutes in, their office manager mentioned she spent four hours a week manually copying purchase order data from PDFs into their accounting system. Nobody had flagged it because it was "just how things worked."
Four hours a week is two hundred hours a year. At her loaded cost, that is roughly $9,000. The automation took us under two weeks to build and cost a fraction of the annual savings. The lead qualification agent is on the roadmap for Q3. But the PDF extractor paid for the entire engagement before we got there.
That is what a backlog-driven strategy looks like. Less impressive in a slide deck. Considerably more impressive on a P&L.
If you want help running this exercise across your team and turning the output into a ranked, costed automation plan, that is exactly what our process is built for.
Need help implementing this?
We build these systems for small businesses and hand you the keys. Book a free discovery call — no sales pressure.
Book a Discovery Call