Key takeaways
- An attendance note is evidence; whichever tool produces it, a fee earner must check and own it.
- Options range from traditional dictation and typing to AI that records, transcribes and summarises meetings.
- Fee-earner time is the main cost driver, so the time a tool saves is worth measuring carefully[1].
- Confidentiality decides which tools are acceptable: free AI tools may not keep what you put in confidential[2].
- Watch AI notes for wrong names, numbers, negatives and speakers; these errors look plausible.
Why attendance notes deserve care
An attendance note records what was said, what advice was given and what the client decided. It is the firm's memory of the matter, the basis for the next step and, when things go wrong, the first document anyone asks for. Communication failures are the most common complaint type brought to the Legal Ombudsman, at 24% in 2025/26, with delay or failure to progress close behind at 22%[3]. A clear, contemporaneous note is often what shows that advice was given and understood.
Tools that speed up note-taking are therefore welcome, but they change who produces the record. The fee earner remains responsible for its accuracy, whether a secretary typed it, speech recognition transcribed it or an AI tool summarised a recording. The SRA is clear that firms remain responsible and accountable for outputs from AI they use[4].
What a good attendance note contains
- Date, start and end time, who attended and how: in person, phone or video.
- Who the fee earner was acting for, and anyone else present, such as a relative or interpreter.
- The client's instructions and the key facts they gave, in their words where it matters.
- The advice given, including risks, options and any time limits explained.
- What the client decided, or that they will decide later and by when.
- Agreed next steps, who does each and the deadline.
- Any concerns about capacity, pressure from others or vulnerability, and what was done about them.
- Time spent, for billing and supervision.
Whatever tool you use, check its output against this list. A beautifully formatted summary that omits the advice given is not an attendance note.
The options compared
| Option | How it works | Main strength | Main risk | Best for |
|---|---|---|---|---|
| Dictation and typing | Fee earner dictates; secretary types | Trusted, human-checked | Turnaround delay; secretary time | Complex advice notes |
| Outsourced typing | Audio sent to a typing service | Overflow capacity | Data leaves the firm; varied quality | Peaks and backlogs |
| Speech recognition | Fee earner speaks; software types live | Immediate, fee earner edits as they go | Recognition errors on names and terms | Fee earners comfortable editing |
| AI transcription of calls or meetings | Recording turned into a full transcript | Complete record of what was said | Consent, storage, speaker mix-ups | Long or contentious meetings |
| AI summary of a recording | Transcript condensed into a draft note | Fast first draft | Omissions, invented detail, misplaced emphasis | Routine calls, with careful review |
Traditional dictation remains the benchmark for quality because a skilled legal secretary catches errors and queries anything unclear. Its weakness is delay: a note dictated on Friday and typed on Tuesday is less useful and less contemporaneous. Speech recognition removes the delay but shifts editing to the fee earner. AI transcription and summarisation go further by working from a recording of the meeting itself, which can produce a fuller record but raises questions about consent, storage and accuracy that the other options do not.
Time is the cost that matters
Most of the cost of an attendance note is the time of the people producing it. ONS provisional figures for 2026 put median gross hourly pay at £31.50 for solicitors and lawyers, £14.95 for legal secretaries and £13.51 for receptionists, before employer costs[1].
Source: Office for National Statistics, 2026[1]
Show the data
| Item | Value |
|---|---|
| Solicitors and lawyers | £31.5 |
| Legal secretaries | £14.95 |
| Receptionists | £13.51 |
Example: a fee earner who spends 20 minutes a day typing or tidying notes loses about 80 hours a year over 240 working days. If a tool cuts that to 8 minutes, the saving is nearly 50 hours of fee-earner time, which is worth more than the hourly pay figure suggests because it can be recorded on matters. Measure the real saving in a trial, including review time, before assuming it.
Confidentiality and data protection
The Legal Ombudsman warns that not all free AI tools are secure or keep what users type confidential[2], and the Law Society's generative AI guidance flags misuse or disclosure of confidential information as a key risk[5]. Before any recording or transcription tool is used on client matters, answer these questions in writing:
- Where are recordings and transcripts processed and stored, and by which sub-processors?
- Is any client audio or text used to train or improve models, and can that be switched off completely?
- How long are recordings kept by the supplier, and can the firm delete them on its own schedule?
- Who at the supplier can access recordings, and is access logged?
- Does the tool work only through the firm's approved accounts, or can staff sign up personally?
Recording a meeting also needs the client's knowledge. Tell clients before recording, explain why and how long the recording is kept, record their agreement in the note and offer a meeting without recording if they prefer. Take extra care with sensitive matters and with third parties present.
Accuracy risks in automated notes
Automated transcripts and summaries fail in ways that look plausible, which makes them harder to catch than a typing error. Check every automated note for:
- Names and spellings, especially of people, companies and places.
- Numbers, dates and amounts, which recognition software often mishears.
- Negatives: a dropped 'not' reverses the advice.
- Speaker attribution: who said what, particularly when a relative speaks for the client.
- Summaries that smooth over doubts, conditions or warnings the fee earner gave.
- Detail that was never said, added by a summarising tool to make the note read well.
Short calls, voicemails and file notes
Most attendance notes are not of long meetings but of five-minute calls: a client confirming instructions, the other side's solicitor proposing a date, a surveyor reporting a problem. These are the notes most often skipped, and the ones automation can help with most, because a short call transcript plus a two-line summary is quick to check.
Set a simple rule for which calls need a note: any call where instructions are given or changed, advice is given, a deadline or date is agreed, or something is said that the firm may need to prove later. Calls that only arrange another call can be logged in the matter history without a full note. Voicemails from clients should be saved or transcribed onto the file when they contain instructions, so the fee earner is not relying on memory of what was said.
Retention and the file
Decide in advance what goes on the file and for how long. Many firms keep the approved note permanently with the matter, keep recordings only for a short, fixed period unless there is a reason to retain them, and record that decision in their retention policy. Whatever you choose, make sure recordings and transcripts held by the supplier follow the same rule, and that deleting them in one place deletes them everywhere.
A review routine that holds up
Review the same day
Read and correct the note while the meeting is fresh. A note reviewed a week later is not much better than none.
Check against the list
Confirm instructions, advice, decisions, next steps and vulnerability concerns are all there.
Mark it as checked
Record that the fee earner reviewed and approved the note, with the date, so the file shows it is not a raw transcript.
Keep or delete the source
Follow a set rule for whether the recording or full transcript is kept on file or deleted after the note is approved.
Sample for supervision
Supervisors review a sample of automated notes each month, comparing them with the recording where one exists.
Choosing between them
Use the option that matches the matter and the fee earner. Many firms end up with a mix: speech recognition for routine file notes, dictation for complex advice, and AI transcription only for long meetings where a full record is valuable and the client has agreed. Use of these tools is spreading: large language models were the most widely used AI technology among UK businesses with ten or more employees in June 2026, used by 18%[6].
Example: a fictional private client team trialled AI summaries for will instruction meetings. Notes were ready the same afternoon, but reviews found two summaries had attributed a relative's comments to the client, which matters when capacity and influence are in question. The team kept AI transcription for those meetings but stopped using automated summaries for them, and fee earners now write the advice section themselves. For support roles affected by these changes, see how AI is changing legal support roles; for introducing the change with the team, see bringing staff with you when introducing AI.
Frequently asked questions
Can we use AI to write attendance notes?
Yes, if the tool is approved for client data, the client knows when a meeting is recorded, and a fee earner reviews and approves every note the same day. The firm remains responsible for its accuracy.
Do we need client consent to record a meeting for transcription?
Tell the client before recording, explain why and how long it is kept, note their agreement and offer to proceed without recording if they prefer.
What errors should we look for in AI transcripts?
Wrong names, numbers and dates, dropped negatives, comments attributed to the wrong speaker, warnings smoothed away in summaries and detail that was never said.
Is speech recognition better than dictation?
It is faster, because text appears immediately, but it moves editing work to the fee earner. Dictation with a skilled secretary still gives the strongest quality check for complex notes.
References
Office for National Statistics (2026). Earnings and hours worked, occupation by four-digit SOC: ASHE Table 14 (2026 provisional) (opens in a new tab) ↑
Legal Ombudsman (2026). AI and Complaints: A Consumer Guide (opens in a new tab) ↑
Legal Ombudsman (2026). Legal Ombudsman 2025/26 annual complaints data and insight (opens in a new tab) ↑
Solicitors Regulation Authority (2023). Risk Outlook report: The use of artificial intelligence in the legal market (opens in a new tab) ↑
The Law Society (2026). Generative AI - the essentials (opens in a new tab) ↑
Office for National Statistics (2026). Artificial intelligence in UK businesses: 2023 to 2026 (opens in a new tab) ↑