Ask a room of fee earners what they would like automated and you get a list of the things that annoy them most. The time sheet. The bundle that takes a day to put together. The client who rings for an update every morning. Those are honest answers, and they are a poor place to start. A firm’s first automation should be chosen by what pays back: what the work costs now, what it would cost to automate and keep running, and whether the time you get back turns into anything. How much it irritates people comes later.
The first project has another job as well. It has to work, be trusted, and show the firm that automation can take real work off people without creating new risk. A project that is clever, rare or full of judgement calls can be a fine piece of software and still lose money and goodwill.
Do the sums before you choose
Start with how the work is done today: who does it, how long it takes, how many times a week, and where the errors come from. You don’t need a consultant for this, and that includes us. A week of tallies on a shared sheet will do.
Put a cost on that time, using what those people cost the firm. Then add up the other side. There’s the licence or the build, the time spent tidying templates and training people, and then the running costs that carry on every year: support, hosting, per-user fees and the minutes a person still spends checking each result.
Next, ask what happens to the time you get back. Payback figures can be too kind here. Time freed for a fee earner is worth money if it goes on chargeable work, or if it lets them take on more fixed-fee matters. A few minutes a day spread across a team may never add up to anything you could point to. Fewer errors and missed dates count as well, though you should be cautious about putting a number on them.
Suppliers’ figures can treat every saved minute as money. We wouldn’t, and you should hold ours to the same standard, because we can also build what we recommend. Write down the assumptions so someone else could check them.
Before paying anyone, look for the cheaper answer. Some steps can be dropped altogether. Some are already covered by a template feature or a workflow setting that’s still switched off. If either is true, you have your first project, and it may cost next to nothing.
Four tests for a first candidate
What’s left after that should pass these four tests, since each makes payback more likely.
It happens often. A task done 30 times a week by several people repays the effort quickly and gets plenty of use as soon as it is live. A task done twice a month will take a long time to prove itself, and any fault in it may sit unnoticed for weeks.
Its rules fit on a page. If a sensible assistant could follow written instructions to do it, a system can follow them too. If the instructions keep saying “it depends”, the work is judgement and belongs with a fee earner.
The information it needs already exists somewhere. Automation that starts by asking someone to type in data the firm already holds in its case management system adds a step and a cost. The better first projects draw on details the firm captured once and reuses.
A person checks the result before it matters. The output goes to someone’s queue to approve, or the system prepares the step and a person makes the decision. That keeps the risk small while trust builds, though the checking time belongs in your figures.
Where that points
Run those tests across a small or mid-sized firm and they point to a handful of kinds of work.
Client care letters and engagement terms pass all four. The letter is drafted from the matter’s details (the client, the fee earner, the supervising partner, the type of work, the estimate) and goes to the fee earner’s queue as a draft to check. Every matter needs one, the content follows the firm’s own template, and a person signs it off.
Then there’s matter opening. A firm can end up collecting the same details at enquiry, at instruction and when the file is opened. Capturing them once and carrying them through is plain, frequent work with clear rules.
In compliance, the administrative side fits. The system can send the ID request, chase the missing document, record what came back and run the search for a conflict check against existing matters. Whether there is a conflict, and whether the source of funds evidence is enough, stays a decision for a person at the firm, and responsibility for compliance stays with the firm.
Reminders and chasers are smaller: undertakings coming due, file reviews, limitation dates, the client who hasn’t returned a signed form. Each may save only a few minutes, but a missed one can be expensive.
Say a residential conveyancing team counts how many client care letters it sends in a week and finds each one is the same template with a handful of fields changed by hand. That beats the bundle that takes a day, because it comes round far more often, the rules are already written in the template, and the fee earner already reads each letter before it goes. The team should still check whether its case management system can merge those fields as it stands, before anyone pays for something new.
What to leave for later
Some work looks like a strong candidate and isn’t, at least for a first project.
Drafting that depends on negotiation, such as heavily tailored commercial agreements, has too many exceptions. Assembly can help later, once the firm trusts its templates, but as a first project it risks turning into a paid argument about clause wording.
Anything spanning several systems with no access between them is a larger piece of work than it looks. Connecting systems is often worth doing, but it adds cost on access and data before anything useful reaches a fee earner, and that pushes the payback a long way out.
Rare work, however painful, gives the system too few chances to pay for itself. The bundle that takes a day once a quarter may never earn back what it costs to automate and maintain.
We would also leave anything where the answer to “who checks this?” is nobody, until the checking step has been designed in.
Check the results afterwards
Keep the tallies, and count again for a week a few months after the change goes live. Without that baseline you can’t tell whether the automation helped, and the project gets judged on feeling. With it, the firm can see what changed and can stop paying for anything that didn’t pay back.
A good first project may also tell you things you didn’t know. Fee earners start to point out the next task that passes the tests, and templates untouched in years get tidied because the system now depends on them. Choosing a second project from there is easier than going back to the original wish list.
If this is your firm
If you have a list of repetitive work and can’t tell which of it would pay back, our AI and technology review follows the work through your firm and sets out the costs for each candidate, including the ones to leave alone.