Automate the highest-volume, lowest-judgment work first, regardless of which department it sits in. Not the task people complain about loudest, not the one that looks most "AI-ready" on the surface, and not whichever department happens to have the most vocal advocate for AI adoption. Volume and judgment are the two variables that actually predict a fast, safe win. Everything else is noise.
Why the obvious answer is usually wrong
Most businesses default to one of two starting points: automate whatever's most annoying, or automate whatever's most visible to leadership. Both feel intuitive. Both usually produce a slow, expensive first project that either doesn't work well or works but doesn't move any number anyone cares about.
The problem is that "annoying" and "visible" aren't the same as "automatable." A task can be miserable and still require enough judgment that automating it half-badly is worse than not automating it at all. Meanwhile, the actual best candidates (repetitive, high-volume, low-ambiguity work) often sit quietly in the background because nobody's complaining about them. They're just... done, over and over, by someone who's stopped noticing how repetitive it is.
A simple way to sequence it
This is also where the distinction we've covered before between automation and augmentation matters: not everything you sequence here needs to remove a human from the task entirely. Score candidate tasks on two axes:
- Volume: how often does this happen? Daily, weekly, a handful of times a year?
- Judgment required: does this need real discretion, relationship context, or handling of ambiguity, or is it the same decision every time?
High volume + low judgment is your starting list. It's where automation is most reliable, fastest to prove out, and least likely to create a visible failure that sours the rest of the organisation on AI.
Everything high-judgment, regardless of volume, should wait. Not forever. Just until you've got a track record of small wins to build trust and process around.
Why sequencing matters more than picking "the right" project
The first automation project a business runs sets the tone for every one after it. Pick something high-volume and low-judgment, and it works cleanly, quickly, and visibly, which buys you permission to tackle harder, more judgment-heavy work later with organisational goodwill intact.
Pick something ambitious and judgment-heavy first because it looked more impressive on a slide, and a rocky rollout there can set the whole initiative back months, regardless of how well the underlying tool actually works.
Where this breaks down without task-level data
As we go into more depth in Task Level Analysis for HR and Business Leaders, the volume-versus-judgment framework only works if you actually know what your teams are spending time on at the task level, not the role level. "Customer service" isn't a task. It's a job title covering a mix of high-volume password resets and high-judgment complaint de-escalation. Automate at the role level and you'll either automate too conservatively (missing the easy wins buried in a "safe" role) or too aggressively (automating the judgment calls inside a role that looks automatable from the outside).
GoFIGR's AI Impact Assessment breaks this down at the task level across your whole business, so the volume-and-judgment sequencing above is based on what people actually do, not a guess dressed up as a strategy.
.jpg)
Helena Turpin spent 20 years in talent and HR innovation where she solved people-related problems using data and technology. She left corporate life to create GoFIGR where she helps mid-sized organizations to develop and retain their people by connecting employee skills and aspirations to internal opportunities like projects, mentorship and learning.
