solo practitioner from Manchester — just joined, here to learn about AI
hi everyone. ACA qualified, been running my own practice for 8 years. about 120 clients, mostly owner-managed businesses and personal tax.
i'll be honest — i'm late to AI. i've been watching from the sidelines while everyone talks about ChatGPT and i finally decided i need to actually engage with it before my clients know more about it than i do.
signed up here because this seems like the one place where accountants are actually discussing AI practically rather than either hyping it or dismissing it.
few things i'd love to learn:
- where to START with AI in a small practice (i use Xero, TaxCalc, and Excel)
- what's actually safe to use AI for vs what's still too risky
- how other solo practitioners are managing the time investment of learning this stuff while still running a practice
looking forward to the conversations.
3 replies
re: "what's safe vs what's risky" — here's my rough framework:
safe to use AI for:
- drafting (emails, memos, letters) — you're reviewing before sending
- research (finding relevant rules, understanding concepts) — you're verifying before relying
- data processing (categorizing, reconciling, formatting) — mechanical tasks with clear right/wrong answers
use with caution:
- tax calculations — verify every number against source data
- compliance deadlines — AI doesn't always have current dates
- anything client-facing that goes out under your name
don't use AI for:
- signing off on positions you haven't independently verified
- citing legal authority without checking it exists (see james's thread about hallucinated revenue rulings)
- anything where an error creates a penalty or liability exposure
welcome emma! you're not late — most of the profession hasn't started yet. you're early relative to the median.
my "where to start" recommendation for a Xero-based practice:
- bank transaction categorization — export uncategorized transactions, feed to Claude with the client's chart of accounts, import the categorized result. immediate time saving, low risk because you're reviewing anyway
- client email drafts — paste the client question, ask AI for a draft response, edit to your voice. saves 50% of the time on routine correspondence
- year-end checklist generation — give AI the client's profile (entity type, industry, jurisdiction) and ask for a year-end close checklist. then customise
start there. all low-risk, all produce visible time savings within the first week.
the "how do you find time to learn" question is the real one. running a practice is already 50+ hours a week. here's what worked for me:
i dedicated Friday afternoons (2-3pm) to "AI experiments." protected time, no client calls. i'd pick ONE task from the week that took too long and try to do it with AI.
most experiments failed. some saved huge amounts of time. after 3 months i had a solid list of "AI works for this" and "AI doesn't work for this" specific to my practice.
the investment pays for itself. but you have to protect the time.
Sign in as a verified accountant to reply.
Sign in