· Dark Oak · 6 min read
Ask a partner at an Irish accountancy firm what AI is doing to their business in 2025 and you’ll get two answers depending on the size. At a four-person practice in Sligo or Letterkenny it’s saving roughly ten hours a week on document handling. At a 200-person firm in Dublin it’s quietly reshuffling the work pyramid — not by removing people, but by changing what the early-career staff actually spend their days on.
We work with professional-services firms across the country — accountancy, legal, management consultancy, surveying — and the gap between the conference-stage talk and the daily reality is wider than usual right now. The genuinely useful adoption is in unglamorous places. The flashy stuff is, for the most part, still a slide deck. This is what we’re seeing actually pay back.
What’s actually being adopted, by use case
There are five workflows where we see firms getting real, repeatable value. None of them are exciting. All of them are saving hours every week.
Document review and extraction
This is the workhorse, and it’s where most firms start. Accountants pulling figures out of bank statements, invoices, and supplier reconciliations. Solicitors extracting key terms from leases, contracts, and bundles. Surveyors pulling specifications out of drawings and project files.
Done well, this doesn’t replace the supervised work the firm is actually paid for. It frees capacity. A trainee accountant who used to spend half a day keying figures from PDFs now spends that time on the review, the queries, and the client conversation — the work that builds judgement and the work the client is paying for. The partner reads a structured summary instead of forty pages, then drills into the bits that matter. The bottleneck shifts up the chain, which is where it should be.
The point isn’t to remove early-career staff from the workflow. It’s to stop spending their hours on data entry that a model handles in seconds with better accuracy than a tired human at 6pm on a Thursday.
Draft-then-review
Letters to clients, internal memos, first-pass advice notes, project scoping documents. The model produces a draft in seconds. The partner or manager reads, edits, signs off.
The interesting thing here is where the time saving actually shows up. It’s not on the draft itself — drafts have always been quick for experienced people. It’s on the revision cycle. When the partner is editing a competent first pass instead of writing from a blank page or correcting an off-target trainee draft, the round-trip collapses. We’ve watched firms cut the time-from-instruction-to-client from three days to one for routine correspondence.
Meeting transcription and structured summarisation
Client meetings, internal stand-ups, discovery calls. Transcribed automatically, summarised into action items, decisions, and follow-ups. Pushed into the CRM or matter-management system.
One thing worth flagging here. Under the EU AI Act’s limited-risk transparency obligations, if an AI is participating in a call — even passively transcribing — the people on the call need to know. Most firms we’ve worked with handle this with a line in the meeting invite and a verbal note at the top of the call. Not difficult, but not optional either.
Knowledge-base agents over the firm’s own files
This is the one that often surprises partners the most. An agent that can answer “have we written something like this before?” against the firm’s accumulated work product — past advice notes, prior client deliverables, internal memos, template libraries.
The honest version of what this saves is inbox and shared-drive archaeology. The question “didn’t we do something similar for that client in Galway in 2022?” used to mean twenty minutes of searching by three different people. Now it’s a sentence and a thirty-second wait. The output still needs reviewing — these systems get things wrong — but the starting point is hours earlier than it was.
Time-tracking and billing narrative drafting
The least glamorous use case on the list, and probably the one with the fastest payback. Turning rough activity notes — “call with client re. lease query, 35 min” — into the kind of client-friendly billing narrative that goes on the invoice. Done at the end of each day rather than in a grim Friday-afternoon catch-up.
Firms tell us this one alone covers the cost of the rest. WIP gets written off less. Bills go out faster. The narratives are more consistent across the firm, which clients notice.
What’s mostly still hype
For balance, here’s what we’d push back on if a vendor is pitching it to you in late 2025.
- End-to-end autonomous client advice. Regulated professions have professional indemnity, statutory duties, and named-individual accountability built into the structure. An agent producing final advice without a human in the loop isn’t a technology problem — it’s a regulatory and liability one. It won’t happen in the timeframes people are quoting.
- “Replacing the firm with an agent.” Clients hire a named partner. They want someone accountable, someone who picks up the phone, someone whose practising certificate is on the line. The agent can do a lot of the work behind that person. It can’t be that person.
- Voice agents handling sensitive client calls. The tech is genuinely improving but accuracy on Irish accents, technical vocabulary, and the tonal judgement a client expects from their accountant or solicitor isn’t there yet. Reception triage and appointment booking — fine. A nuanced conversation about a tax position or a family-law matter — not yet, and we’d be cautious of anyone telling you otherwise.
- Anything pitched as an “AI partner.” The regulatory ceiling is real and not going anywhere soon. Be skeptical of marketing that implies otherwise.
The honest answer to the fear question
The question we get asked privately, usually over coffee after the formal meeting, is some version of: are we going to need fewer trainees?
The honest answer is that the trainee and associate structure of Irish professional-services firms will change. The change is slower and more uneven than the LinkedIn headlines suggest, but it’s real. Some firms will end up with smaller intake cohorts. Others will hire the same numbers but have them doing different work earlier in their careers.
Firms that pretend nothing will change will struggle, because their cost base will sit above competitors who’ve adopted the workhorse use cases above. Firms that try to replace early-career people wholesale will struggle harder, because they’ll lose the pipeline of future managers and partners — and clients will feel the drop in service quality long before the firm does.
The ones we see doing well are the ones picking specific workflows, measuring the hours saved, and reallocating those hours into client work the firm is genuinely paid for. That’s the entire story. There’s no clever framework underneath it.
Where to start if you’re a 5-50 person firm
Most firms in this size band don’t need a custom build. They need someone to spend a focused week or two looking at how the firm actually works, identifying the two or three workflows with the largest weekly time cost, and figuring out which of those is best solved with configuration of existing tools versus a small custom agent.
That’s what our Readiness Sprints are for — fixed-price from €1,500, a couple of weeks, output is a prioritised plan you can act on with or without us. From there, most firms pick one workflow to tackle first. Sometimes that’s configuration of a tool they already pay for. Sometimes it’s a small custom agent — our AI and automation builds start at €4,000, fixed price, with EU data residency baked in because for Irish professional services that isn’t optional.
The firms we’ve worked with that have got the most out of this are the ones that picked one workflow, measured what it actually saved, and only then moved to the next. The ones that tried to do everything at once spent a lot and got tangled. Start small. Measure. Move on.
If you’re running a firm and trying to work out where to start, or if you’ve already tried something that didn’t quite land and want a second opinion, get in touch. We’ll tell you honestly whether what you’re considering is worth doing — and if it isn’t, what would be.