Industries / Technology / Software and SaaS / Field service software
Customer updates
Booking the visit, telling the customer when the technician will arrive, and keeping them informed when plans change.
6
Milestones achieved, from 1997 to 2026.
2
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1997
Appointments kept by law
Australia's Customer Service Guarantee, in place since 1997, sets maximum times to connect and repair fixed telephone services, "including associated appointment-keeping arrangements", and benchmarks set by the minister require 90% compliance. At Telstra today a missed appointment earns a home customer A$14.52 and a business A$24.20, and a window of up to four hours carries 15 minutes' grace.
Credit
Australian Government
2014
Tracking the technician
In November 2014 Comcast began testing an app near Boston that alerted customers when the technician was about 30 minutes away, showed the technician on a map and let them rate the visit; the trade press called it "an Uber-like model". Dish Network launched a nationwide tracker, My Tech, in September 2015, and Time Warner Cable launched TechTracker, with the technician's name, ID and photo, in October 2015.
Credit
Comcast; Dish Network; Time Warner Cable
2015
Twenty dollars for lateness
Four-hour arrival windows were common for cable installations in 2011, and Comcast said then that "most customers are fine if a window is two hours or less." In May 2015 its cable chief, Neil Smit, promised that if a technician was even one minute late, the customer would get an automatic $20 credit. Comcast's guarantee page now promises "convenient appointment windows" and lists no credit.
Credit
Comcast
2016
Missed by contractors
Britain's energy regulator, Ofgem, said E.ON paid £3.1 million after agents working for it missed appointments and E.ON failed to pay the compensation the rules require. In 2017 British Gas paid £1.1 million after third-party agents missed appointments and it failed to compensate about 12,000 customers, most of them businesses.
LessonThe duty to compensate stays with the firm that made the appointment, even when a contractor misses it.
Credit
Ofgem
2019
Paid without asking
From 1 April 2019 British phone and broadband providers in Ofcom's automatic compensation scheme, among them BT, Sky, TalkTalk and Virgin Media, pay customers without being asked when an engineer misses an appointment or a repair or installation runs late. Ofcom said in November 2017 that compensation was then paid "in around one in seven cases (15%)", and set £25 per missed appointment; BT now pays £32.31.
Credit
Ofcom
2026
Half of appointments kept
Australia's competition regulator, the ACCC, reported that NBN Co, which runs the national broadband network, "met the appointment time for appointments that were not rescheduled in just over half (around 56%)" of all appointments over the 12 months to December 2025.
Credit
ACCC
Watch for
Agents that move the appointment
Salesforce says its scheduling agents can now handle appointment booking on their own and "proactively reach out to renegotiate times before a delay becomes an issue", and that an unnamed client saved about 20 minutes per appointment.
Achieved whena named customer or an outside party publishes on-time rates and customer complaints before and after, with its method.
Claimed by
Salesforce, April 2026
After-hours bookings by AI
Housecall Pro says that "26% of all jobs booked by CSR AI come in after hours", CSR AI being its AI that answers a firm's calls and texts, and cites only "Housecall Pro data".
Achieved whenan outside party checks a sample of the bookings, including whether urgent calls reached a person. The plumbing and electrical contracting files propose such a test for AI call-takers.
Claimed by
Housecall Pro, August 2026
Open
01
A shared on-time score
A published definition of how to score arrival promises (how wide the window was, whether the technician arrived inside it, how far the predicted arrival time was off in minutes, how much notice the customer had of a change) and an open tool that computes the score from any field service app's job export. Then a first public table for firms that agree to take part.
The briefCredit · unclaimed02
Delay notices in many languages
A set of 200 situations a dispatcher has to explain to a customer (a technician running late, a part on back order, a job moved to tomorrow) and a scoring of AI-written messages for each, in at least ten languages, by native speakers: whether the time, name and reason are right, and whether the message is clear.
The briefCredit · unclaimed
Open 01 · the brief
A shared on-time score
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A published definition of how to score arrival promises (how wide the window was, whether the technician arrived inside it, how far the predicted arrival time was off in minutes, how much notice the customer had of a change) and an open tool that computes the score from any field service app's job export. Then a first public table for firms that agree to take part.
Why it is still open
Regular outside measures exist where regulators require them: Australia's ACCC for the NBN and the Customer Service Guarantee for fixed phones, and Ofcom's compensation scheme in Britain. This research found no shared score for other trades, and no public test of how far predicted arrival times are off.
Built with what exists
job exports with booked windows and arrival times from GPS or check-in; Python; the Customer Service Guarantee's grace-period rules as a model.
First step
Write the definitions on one page and score one firm's last month of jobs.
Done looks like
10 firms in at least two countries scored over three months, with definitions and code published, and an outside party recomputing one firm's score from its raw data to within one percentage point.
Open 02 · the brief
Delay notices in many languages
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A set of 200 situations a dispatcher has to explain to a customer (a technician running late, a part on back order, a job moved to tomorrow) and a scoring of AI-written messages for each, in at least ten languages, by native speakers: whether the time, name and reason are right, and whether the message is clear.
Why it is still open
Vendors now let AI agents contact customers about delays (see Watch for). τ²-Bench, a public test of customer-service agents, includes telecom troubleshooting but not appointment or delay messages, and Microsoft tests its work order summaries in English only. This research found no test of such messages.
Built with what exists
current AI models from several makers; scenario cards written by working dispatchers; native-speaker raters; a simple scoring sheet.
First step
Write 20 scenario cards and score messages from three models in English and one other language.
Done looks like
200 scenarios in ten languages, each message rated by two native speakers, with error rates by model and language published, and a second group of raters agreeing on a 10% sample.