What problem does it solve?
Every business that books time with customers loses some of it. Patients miss appointments, engineers arrive at empty houses, delivery drivers carry parcels back to the depot and restaurant tables stay empty. The cost is not only the lost slot: it is the other customer who could have had it, and the repeat visit that now has to be arranged.
Many organizations already send reminders, but classic reminders are one way broadcasts at a fixed moment. They cannot answer "can I come an hour later", they do not know which customers are likely to miss, and a cancellation made the evening before usually leaves the slot empty. Human confirmation calls allow a real conversation but are costly to make for every booking.
- The UK government says sending reminders has been shown to reduce missed appointments by up to 80%, and that NHS trusts report better results when communication with the patient is two way.Power to patients as government sets out plan to cut waiting lists (2025)
- NHS England reports that eight million (6.4%) of 124.5 million outpatient appointments in England in the previous year were not attended.NHS AI expansion to help tackle missed appointments and improve waiting times (2024)
How does it work?
- The booking system triggers the contact. Reminders and confirmations start from an existing booking, order or reservation, at times set per service. The agent does not choose who to contact for commercial reasons.
- Prioritise where it matters. A risk score can decide who gets an extra reminder, a call instead of a text, or an offer of help, as Sheffield Children's did in its AI Predictor pilot.
- Open with who and why. The agent names the organization, says it is an automated assistant and states the booking it is about, without revealing sensitive details before the person is verified.
- Let the customer act. The customer can confirm, cancel, move to another offered slot or ask a practical question (address, parking, preparation, delivery window) in the same conversation.
- Write back and reuse. Confirmations, cancellations and new times go straight back to the booking system, and freed slots are offered to the next customer on the waiting list.
- Escalate what is not routine. Complaints, distress, clinical questions and customers who ask for a person are handed to staff with the conversation attached.
- Audience
- Customer facing
- Autonomy
- Supervised agent
- Adoption
- Early adopters
- Channels
- Phone and voice, SMS, WhatsApp, Email
What is it worth?
Benchmarks are computed from the public deployments below: one data point per organization per KPI, with who made each claim.
| KPI | Median | Reported range | Data points | Claimed by |
|---|---|---|---|---|
| Interactions handled | Not pooled | at least 160,000 | 1 | 1 organization |
Value drivers: Lower cost to serve, Customer experience, Speed and cycle time, Inclusion and access.
Indicative value
A service organization with 200,000 booked appointments or visits a year
EUR 40,000 to EUR 480,000
Value of bookings kept or refilled per year
How this is calculated
Formula: bookings * missedRate * reduction * valuePerBooking. The low scenario uses every low input, the high scenario every high input.
| Input | Low | High | Basis |
|---|---|---|---|
| Booked appointments or visits per year bookings, bookings per year | 200,000 | 200,000 | The reference organization. |
| Share of bookings missed without notice today missedRate, fraction of bookings | 0.05 | 0.1 | Editorial assumption. NHS England reports 6.4% for outpatient appointments; replace with your own rate. |
| Share of missed bookings avoided or refilled because of the agent reduction, fraction of missed bookings | 0.1 | 0.2 | Editorial assumption, set at or below the NHS pilots on this page. Sheffield Children's expected 8,581 missed appointments against a benchmark rate, not a control group, and recorded just under 6,500, about a 24% reduction. UHCW's move from 10% to 4% came from reminder timing found through process mining, not from an agent. Measure against a control group. |
| Value of a booking that is kept or refilled valuePerBooking, EUR per booking | 40 | 120 | Editorial assumption covering the margin or cost of a wasted slot or visit. Replace with your own figure. |
What it leaves out: Counts kept and refilled bookings only. It leaves out staff time saved on manual confirmation calls, the cost of messages, calls and the AI, and any annoyance from contact that customers did not want.
Who already uses it?
4 public deployments, strongest evidence first. Grades: A regulator or audit, B the organization itself, C vendor case study, D anonymous or estimate.
University Hospitals Coventry and Warwickshire NHS Trust
United Kingdom · Healthcare · 2024
University Hospitals Coventry and Warwickshire used AI based process mining with IBM and Celonis to study missed appointments, which were more common among patients with high deprivation scores. It found a spike in last minute cancellations after two SMS reminders and moved to a reminder 14 days before the appointment with a second one four days before, so patients could cancel early and the slot could be rebooked. NHS England reports that missed appointments in this subset of patients fell from 10% to 4%. The AI analysed the process; the reminders themselves are standard text messages.
No outcome disclosed.
WellSpan Health
United States · Healthcare · 2024
WellSpan Health started with a Hippocratic AI voice agent that phones patients to close gaps in colorectal cancer screening, in English and Spanish, and supports low risk patients before and after a scheduled colonoscopy, with transcripts sent to clinicians and live transfer to a human where needed. By 2026 the agent, Ana, answered inbound calls and scheduled primary care appointments, and WellSpan announced an expanded partnership whose first new workflows call patients who missed imaging appointments. WellSpan reports that Ana manages more than 160,000 patient calls a month.
- Interactions handled: at least 160,000, per month, patient calls
"Ana, WellSpan’s generative AI agent developed in partnership with Hippocratic AI, currently manages more than 160,000 patient calls per month, engaging in more than 7,000 hours of conversation."
Claimed by: organization
Sheffield Children's NHS Foundation Trust
United Kingdom · Healthcare · 2023
Sheffield Children's piloted an AI Predictor developed by Alder Hey Innovation that estimates which children are likely to miss ("was not brought") an appointment, using markers that include health inequalities. Families with a predicted risk of 50% or more received an extra text reminder with an offer of support the day before; families at 85% or more were contacted and offered funded transport or a rebooking. NHS England reports that 53,800 texts were sent in the first 12 months, that recorded non attendance came in well below the expected benchmark (almost 200 more attended appointments a month), and that in a 13 week period 152 families had transport arranged and 129 appointments were rebooked.
No outcome disclosed.
US Department of Veterans Affairs, Veterans Benefits Administration
United States · Government and public sector · 2020
The Veteran Readiness and Employment service of the Veterans Benefits Administration used e-VA, a digital assistant acquired from SaraWorks, to take routine follow up off vocational rehabilitation counselors. It created reminders for appointments, grades and receipts and sent routine messages by text or email; participants could reply to schedule and reschedule appointments, respond to reminders and send documents. The 2024 federal AI inventory lists it in operation since June 2020 and classifies it as both rights and safety impacting; the 2025 inventory lists it as retired. No results were published.
No outcome disclosed.
How do you implement it?
A model agnostic playbook: what to prepare, the order to build in, and what goes wrong.
Data you need
- Bookings with time, location, service type and contact details
- Available alternative slots and a waiting list, if freed slots are to be reused
- Channel preferences, consent and opt outs per customer
- Historic attendance data if contacts are prioritised by risk
Systems to integrate
- Booking, scheduling or order management system (read and write)
- Telephony and messaging (voice, SMS, WhatsApp, email)
- Waiting list or capacity planning
- CRM or case management for follow up
- Contact centre for handover
Complexity: Medium
The dialogue is short. The effort goes into a clean trigger from the booking system, a reliable write back of confirmations and changes, contact rules per channel and market, and verifying the person before details are shared.
- 1
Start with the reminders you already send
Make the existing reminder two way before adding new ones: let the customer confirm, cancel or move it by replying. This alone turns late cancellations into slots you can reuse.
- 2
Test the timing
Timing matters as much as wording. University Hospitals Coventry and Warwickshire found a spike in last minute cancellations after two text reminders, and found that reminders 14 days and four days ahead worked best because patients cancelled early enough for the slot to be rebooked.
- 3
Add risk based escalation
Use a missed booking score to decide who gets an extra message, a call or practical help, and keep a control group so you can see the effect.
- 4
Connect cancellations to the waiting list
A reminder that produces a cancellation is only valuable if someone else gets the slot. Automate the offer to the next suitable customer.
- 5
Keep marketing out
Do not add offers or upsell to service reminders. It changes the legal basis for the contact and the trust customers place in the channel.
Guardrails
- Contact only about an existing booking, order or reservation, never for sales
- Disclose that the caller or sender is an automated assistant and name the organization
- Verify the person before sharing sensitive details such as a clinic or diagnosis
- Honour opt outs and quiet hours on every channel
- Never ask for passwords, card details or payment in a reminder
KPIs to instrument
- Missed booking rate against a control group
- Share of cancellations made early enough to refill, and refill rate
- Confirmation, cancellation and rebooking rates per reminder
- Opt out and complaint rate per reminder type
- Reach and outcomes by language, age and deprivation band
Human in the loop
Service owners approve every reminder script, timing and channel. Staff take over complaints, distress, clinical or safety questions and anyone who asks for a person, and review a sample of conversations and outcomes each month, with attention to groups that miss the most bookings.
Common failure modes
- Reminders that look like scams
- Unexpected calls and texts are what fraudsters imitate. Name the organization, never ask for secrets, and point to known channels.
- Late cancellations with no refill
- The reminder works but the slot stays empty. Time reminders so cancellations come early and connect them to a waiting list.
- Privacy leaks in the message
- A reminder names a sensitive clinic or service to whoever reads the phone. Keep content minimal until the person is verified.
- Service outreach turning into marketing
- Offers creep into reminders and the contact now needs marketing consent. Keep scripts service only and review them.
What are the risks and rules?
EU AI Act
Depends on design
People must be told they are interacting with an AI system, and synthetic voice or text must be identifiable as such (Article 50). Reminding people of existing bookings and disclosure alone are limited risk. A missed appointment score used by or for a public authority to grant, reduce, revoke or reclaim access to healthcare or other essential public assistance and services, for example deciding who is offered funded transport, can fall within Annex III point 5(a), and profiling of natural persons within Annex III rules out the Article 6(3) exemption. Using the score only to decide who gets extra reminders or support does not by itself place it outside Annex III when that support is itself the assistance being granted.
Rules that apply
Guidance
- Declaratory ruling on AI generated voices under the TCPA (FCC 24-17) (Federal Communications Commission, North America). AI generated voices count as artificial or prerecorded voice under the TCPA, so US reminder calls made with AI need the consent the TCPA requires for such calls.
- Guide to PECR: electronic and telephone marketing (Information Commissioner's Office, Europe). UK rules on marketing by phone, text and email. Routine customer service messages about a current contract or past purchase, such as delivery arrangements, do not count as direct marketing; unsolicited marketing often needs specific consent.
Controls to put in place
- Documented purpose and legal basis for each reminder type
- Opt out, quiet hour and frequency checks logged per contact
- Approved scripts per reminder type with an accountable owner
- Audit trail of every contact and every change written back to the booking system
- Monitoring of complaints, opt outs and outcomes by group
Frequently asked questions
- How is this different from proactive outreach and activation?
- Proactive outreach contacts customers about something they have not done yet, such as activating a card or accepting an offer, and often needs marketing consent. Reminder and confirmation outreach is about something the customer already booked or ordered and exists to make it happen as planned.
- Do AI targeted reminders reduce no shows?
- NHS pilots suggest so. In the first 12 months of a pilot at Sheffield Children's, an AI Predictor sent 53,800 extra text reminders to families at high risk of missing appointments, and NHS England reports almost 200 more attended appointments a month against the benchmark. Results without a control group should be read with care.
- Should reminders be calls or messages?
- Messages for most people, because they are cheap and let the customer reply when it suits them. Calls suit high risk bookings and people who do not use messaging. WellSpan Health first used its AI voice agent to call patients about colorectal cancer screening, and has announced outreach to patients who missed imaging appointments as a next workflow.
- Are AI reminder calls legal?
- It depends on the market and the purpose. In the US, AI generated voices fall under the TCPA's rules for artificial voices; in the UK, PECR does not count routine customer service messages as direct marketing. Keep reminders strictly about the booking and record the legal basis.
How to cite this page
Blits.ai AI Use Case Library, "AI agent for outbound reminders and confirmations by voice and messaging", last verified 26 September 2026, https://www.blits.ai/ai-use-cases/outbound-reminder-and-confirmation-agent. Licensed under CC BY 4.0. Method: how we verify use cases.
Changelog
- 27 September 2026: First published