My company knows AI could help but we have no idea where to start. How do we identify the highest-value opportunities?
Start with the work, not the technology. List the tasks in your business that consume the most hours, then score each one on four signals: it happens often, it follows rules, the information it needs already exists somewhere, and someone hates doing it. The tasks that score high on all four are your highest-value opportunities. Pick one, build it, and let the hours it saves pay for the next.
By Cade Dannels, Founder, Daita Dynamics. Updated .
Why "where do we start" is the right question
Most businesses that struggle with AI did not fail at technology. They failed at selection: they started with a tool someone saw a demo of, or with the most exciting idea rather than the most valuable one. Choosing the first project well matters more than choosing the first tool, because a first project that saves real hours builds the confidence and the budget for everything after it.
Step one: list the work
Sit down with the people who do the operational work, not only their managers, and ask three questions:
- What do you do every week that feels like the same thing over and over?
- What takes far longer than it should?
- What do you dread?
You will get a list of ten to twenty tasks. Typical answers at growing B2B businesses: assembling month-end reports for clients, onboarding new customers, chasing documents, re-keying data from one system into another, triaging a shared inbox, scheduling, preparing proposals from the same building blocks, and answering the same internal questions repeatedly.
Step two: score each task on four signals
For each task, score one to five on these four signals.
Volume. How often does it happen, and how many hours a week does it consume across the team? A task that eats ten hours a week is worth far more than one that takes an afternoon a quarter.
Rules. Could you write down how to do it? Tasks that follow a consistent process, even a complicated one, automate well. Tasks that depend on judgement every time do not, yet.
Data already exists. Does the information the task needs live somewhere a system could read: a CRM, an accounting tool, a spreadsheet, a shared drive? If it lives only in someone's head, the task needs an organising step before it can be automated.
Someone hates it. This one is not sentimental. Tasks people dislike get done late, inconsistently and with errors, and the people doing them will champion the automation instead of resisting it.
Step three: rank and choose
Add the scores. The top of the list is your candidate. Before committing, check two things. First, is there a clear owner inside your business who will run the system once it exists? Second, can you state the outcome in one sentence, such as "client reports go out on the second business day instead of the tenth"? If both are yes, that is your first project.
Resist the temptation to start three things at once. One working system that saves real time is worth more than three half-built ones.
Step four: let the first win fund the second
The first automation should be chosen partly for how it sets up the next one. Organising your client data to automate reporting also makes onboarding automation easier. Structuring your knowledge for an internal question-answering system also improves proposal assembly. This is why the order matters: organise the knowledge, automate the workflows on top of it, then build systems that learn from the results.
Examples by kind of business
- Professional services firms: month-end or quarter-end client reporting that pulls from several systems into a formatted deliverable.
- Fractional executives and consultants: meeting notes and follow-ups, so every call produces actions without a manual write-up.
- Commercial real estate and finance teams: extracting and normalising data from documents that arrive in inconsistent formats.
- Home services and field businesses: scheduling and dispatch that respond to changes without a person re-planning the day by hand.
- Any business with a shared inbox: triage that routes, tags and drafts replies so people handle exceptions rather than everything.
What to do if nothing scores high
Sometimes the honest result is that no task scores well yet, usually because the information is scattered. That is a finding, not a failure. Your first project is then an organising project: getting the knowledge your business runs on into a structured, searchable form. It is less glamorous and it is what makes every later project possible.
Where Daita Dynamics fits
This is the exercise Daita Dynamics runs in its free AI Opportunity Assessment. Founded in Denver in 2024 and working remotely with growing businesses across the United States, we sit with the people who do the work, score the tasks, and hand back a ranked list with a recommendation. If the right first step is a simple no-code fix your team can do itself, we say so. If you want to see the method applied fast, the 24-hour challenge builds one working automation for a single task in a day.
Frequently asked questions
- Should we start with the biggest problem or the easiest one?
Start with the one that scores highest on all four signals, which is usually neither the biggest nor the easiest. A high-volume, rule-based task with data already available is the sweet spot: meaningful hours saved with manageable risk.
- How much time does the scoring exercise take?
A half day with the right people. The hard part is not the scoring; it is getting the people who actually do the work into the room.
- Do we need a consultant to do this?
No. The method above is enough to produce a good first list. Outside help speeds it up, adds an honest view of how ready your data is, and matters most when it comes time to build.
- What if the most valuable task needs judgement every time?
Automate the parts around it. Most judgement-heavy tasks are wrapped in preparation and follow-up that is entirely rule-based. Take those off the person's plate and leave them the decision.