Key takeaways
- Over 30% of GP time goes on indirect care such as letters and results[1]; trained admin staff can take much of the letter load safely.
- A written protocol decides which letters admin staff code and file and which must go to a GP, with a named GP lead.
- Audit a sample every month and track errors; the goal is fewer letters reaching GPs without any loss of safety.
- Under 2026 MHRA guidance, AI voice tools that only draft or suggest codes for review are not medical devices, but clinicians stay responsible[2].
The letter problem
Every GP practice receives a steady stream of clinical correspondence: discharge summaries, outpatient letters, test results from other providers, out-of-hours reports, screening results and notifications. Each one needs to be read, matched to the right patient, coded where relevant, acted on if needed and filed. If every letter goes to a GP, it competes with consultations for the most limited resource in the building.
NHS England estimated in its 2023 access recovery plan that over 30% of GP time is spent on indirect patient care, including paperwork such as referral letters, fit notes and medical certification, and analysing and responding to test results[1]. The same plan reported that administrative tasks outside a consultation, measured by entries to medical records, were up 50% since 2019[1]. Meanwhile the average number of patients per full-time-equivalent GP in England reached 2,143 in September 2026[3].
Over 30%
of GP time spent on indirect patient care, including letters and results[1]
What workflow optimisation means
Workflow optimisation, sometimes called correspondence management or document workflow, is a simple idea. Trained administrative staff read incoming letters, code the relevant information, action routine requests and file them, following a protocol agreed by the GPs. Only letters that need a clinical decision go to a GP, and when they do, they arrive with the key action highlighted.
Done well, it frees GP time without reducing safety, because the protocol is explicit, staff are trained and signed off, and the work is audited. Done badly, it moves risk onto untrained staff with no oversight. The difference lies in the protocol, the training and the audit.
Source: British Medical Association, 2025[4]
Show the data
| Item | Value |
|---|---|
| GPs | 22% |
| All doctors | 12% |
Workload pressure is the reason most practices start. In the GMC's Workplace Experiences 2025 report, 22% of GPs said they were unable to cope with their workload at least once a day, compared with 12% of all doctors[4].
A protocol by letter type
| Letter type | Admin action | Send to GP when |
|---|---|---|
| Outpatient letter, no actions for GP | Code diagnoses and procedures, file | Any new diagnosis of concern, or the letter is unclear |
| Outpatient letter with a GP action | Code, create task for the action | The action is a medication change, a referral or a clinical judgement |
| Discharge summary | Code, flag medication changes for the pharmacist or GP | Always for medication changes; follow-up requests |
| Did-not-attend notification | Code, check safeguarding flags | Child or vulnerable adult, or repeated non-attendance |
| Screening result, normal | Code and file per programme rules | Never routinely, unless the protocol says otherwise |
| Screening result, abnormal | Code | Always |
| Out-of-hours report | Code, file | Follow-up requested, safeguarding, or prescribing |
| Anything unclear | Do not guess | Always |
Governance that keeps it safe
- A named GP lead owns the protocol, signs off staff competency and reviews audits.
- The protocol is written, versioned and dated, and reviewed at least once a year or after any incident.
- Staff complete training and a supervised period before working independently, with competency recorded.
- There is a clear rule that anything unclear goes to a clinician, and staff are never criticised for escalating.
- Workflow is part of the practice's significant event process, so learning from errors is shared.
- The protocol is covered in induction for new GPs and locums, so they know what they will and will not see.
What to code, and how
Coding is where letters turn into usable records. Agree a short coding guide alongside the protocol: which diagnoses, procedures, medication changes and results should be coded, using which SNOMED CT concepts, and what should be left as free text. Over-coding clutters records; under-coding means searches, recalls and contract reporting miss patients.
- Code new diagnoses and significant procedures from secondary care letters.
- Code results that drive recalls or registers, such as screening outcomes.
- Record medication changes as tasks for clinical review rather than changing prescriptions directly.
- Use templates or quick-picks for the most common codes to reduce variation.
- Avoid coding suspected diagnoses as confirmed; follow the protocol on uncertain wording.
Who should do the work
Letter handling suits people who are careful, curious about medical language and comfortable asking questions. Many practices draw from experienced receptionists or medical secretaries who already know the patients and the clinical system. Whoever you choose, protect their time. Correspondence work squeezed between phone calls and the front desk is where mistakes happen.
- Give workflow staff a quiet space and fixed hours away from the phones.
- Make sure at least two people are trained, so annual leave and sickness do not create a backlog.
- Provide a medical terminology course and access to a clinician for quick questions during each session.
- Recognise the extra responsibility in job descriptions and pay bands, which helps retention.
- Hold a short monthly meeting between workflow staff and the GP lead to discuss tricky letters and updates to the protocol.
If the PCN employs clinical pharmacists, involve them in the protocol for medication changes in discharge summaries and outpatient letters. They are often better placed than GPs to reconcile medicines quickly.
Setting it up step by step
Measure the baseline
For two weeks, count letters by type and how many each GP handles. Note how long the backlog runs.
Draft the protocol
The GP lead and an experienced administrator draft rules for each letter type, starting cautiously.
Train and supervise
Staff learn the protocol and the coding guide, then work under supervision with every letter checked.
Go live in stages
Start with the lowest-risk letter types and add others as audit results allow.
Audit monthly
Review a sample of letters handled by admin staff for correct coding, filing and escalation.
Review and expand
Every quarter, review error rates, GP letter volumes and any incidents, and adjust the protocol.
Auditing without drowning in it
A practical audit samples a fixed number of letters per administrator each month, chosen at random, and checks four things: the right patient, the right codes, the right action and the right escalation. Record errors by type and severity. A rising error rate in one letter type is a signal to retrain or to move that type back to clinicians. Share results with the team and with GPs, because confidence in the system is what lets GPs stop checking everything themselves.
Measure what matters: letters reaching GPs, time from receipt to filing, and audited error rate. If GP letter volume falls and errors stay low, the system is working.
Where technology helps
Document management tools route letters into queues, highlight key content and track who handled what. Newer AI tools can read letters, suggest codes and draft summaries. The government's 10 Year Health Plan commits to rolling out ambient voice technology, digital triage and the single patient record to cut clinical admin within two years[5], and estimates that saving just 90 seconds per GP appointment would generate over 2,000 full-time-equivalent GPs' worth of capacity[6].
MHRA guidance on ambient voice technology, published in July 2026, says products intended solely for tasks such as summarising, drafting letters or suggesting clinical codes for a clinician to review are not regulated as medical devices, and that clinicians remain responsible for reviewing and verifying AI-generated outputs[2]. In a letter workflow, that means AI suggestions sit inside the same protocol: a trained person confirms every code, and anything clinical still reaches a clinician. Among GPs already using AI, 57% use it for clinical documentation and note-taking and 44% for administrative tasks[7].
Pushing back on avoidable work
Some letters create work that should not land in general practice at all, such as requests for the GP to chase results ordered by the hospital or to prescribe a medicine the hospital should supply. The access recovery plan required ICBs to report progress on improving the interface between hospitals and general practice[1]. Keep a simple log of these letters and share it with your PCN and LMC; patterns are easier to fix than individual cases.
Example: a phased introduction
Example: a fictional practice with six GPs started by letting two trained administrators code and file outpatient letters with no GP actions. After two months of audits with very few errors, the practice added did-not-attend notifications and normal screening results. GPs reported spending noticeably less time on correspondence, and the backlog of unfiled letters, which had run to several days, cleared within the week.
Frequently asked questions
Is it safe for admin staff to code clinical letters?
Yes, within a written protocol agreed by GPs, with training, signed-off competency, a rule that unclear letters go to a clinician, and regular audit.
Which letters should always go to a GP?
Typically abnormal results, medication changes, new serious diagnoses, safeguarding concerns and anything unclear. Your protocol should list them.
Can AI code our letters automatically?
AI can suggest codes, but a trained person should confirm them. MHRA guidance says clinicians remain responsible for reviewing AI-generated outputs[2].
How much GP time can workflow optimisation save?
It depends on letter volumes and how much of the protocol you adopt. Measure GP letter volumes before and after to see your own result.
References
NHS England and DHSC (2023). Delivery plan for recovering access to primary care (opens in a new tab) ↑
GOV.UK (Medicines and Healthcare products Regulatory Agency) (2026). MHRA clarifies regulatory status of ambient voice technologies used in the NHS (opens in a new tab) ↑
British Medical Association (2026). Pressures in general practice data analysis (opens in a new tab) ↑
British Medical Association (2025). Pressures in general practice data analysis (opens in a new tab) ↑
GOV.UK (Department of Health and Social Care) (2025). Fit for the future: 10 Year Health Plan for England (opens in a new tab) ↑
GOV.UK (Department of Health and Social Care) (2025). Fit for the future: 10 Year Health Plan for England (opens in a new tab) ↑
Nuffield Trust (2025). How are GPs using AI? Insights from the front line (opens in a new tab) ↑