Human-centered design systems and product experiences
Core services
01UI/UX Design02Web Development03Software Development042D/3D Experiences05App Development06AI Agents & Automation07Website DesignTechnical services
Frontend Engineering
React DevelopmentNext.js DevelopmentWeb ApplicationsEnterprise DashboardsSaaS PlatformsBackend & Infrastructure
API DevelopmentCloud InfrastructureDevOps & CI/CDBackend SystemsCustom SoftwareIndustries we serve
Technology & Software
AI Telecom Troubleshooting Agent Development
Outage checked first, then the steps
An agent for "my internet is not working".
An agent for "my internet is not working". It checks for a known outage first, walks the customer through your approved diagnostic steps one at a time, reads the line's status where your systems allow, and books an engineer or routes the case when the steps do not fix it.
A broadband fault call follows a script. Is the router on, which lights are showing, have you restarted it. The customer has already done all of this. Twenty minutes in, it turns out there is an outage in their area that the network team knew about an hour ago. The steps are sensible. Running them in the wrong order, by voice, for every caller, is what makes the experience bad.
This page describes a workflow we propose to build as a custom project. It does not exist as software yet. It would reuse the runbook-driven diagnosis that runs in our IT helpdesk demonstration on test data: questions from an approved runbook, one permitted fix after consent, and escalation with a trace when it does not work. Line tests, outage data and engineer booking are not built.
Six jobs it carries through
01Outage check first
Checks for a known problem before anything else.
- Known outages at the address
- Planned maintenance
- The published estimate
- No diagnostics during an outage
02Account and service check
Rules out non-technical causes.
- Identity confirmed
- Service active or suspended
- Recent changes to the account
- Routed to billing where relevant
03Guided diagnostics
Follows your approved steps.
- Asks what was already tried
- One step at a time
- Photos of the equipment's lights
- Stops when the answer points elsewhere
04Line status and resets
Uses what your systems permit.
- Reads line and equipment status
- A remote reset from your permitted list
- Only with the customer's consent
- The result reported plainly
05Engineer booking
Books a visit when one is needed.
- Only after the runbook says so
- From real engineer availability
- With the diagnostics attached
- Access instructions collected
06Case continuity
Keeps one case for the fault.
- Everything tried so far
- Recognises a repeat contact
- Escalates a recurring fault
- A summary for the next person
From the First Signal to the Record
An example of one case, in 6 steps. The same steps play in the picture at the top of this page.
Customer writes
A customer reports that their home broadband stopped an hour ago.
Outage checked
There is no known outage at their address and no planned work.
Account checked
The service is active and nothing changed on the account.
Diagnostics run
They have already restarted the router. A photo shows the line light off.
Reset tried
With consent, a remote line reset from your permitted list is run. The light stays off.
Engineer booked
An engineer visit is booked from real availability, with the diagnostics attached.
Walk One Real Case Through It
Thirty minutes, one case from your own process. We map it against the workflow on this page and tell you plainly what can be reused from our demonstrations and what has to be built for your systems.
Six Things That Slow This Down
The script before the outage check
Twenty minutes of steps during a known network fault.
Steps the customer already tried
Nobody asks what they did before calling.
Lights described by phone
"The second one is sort of orange."
Repeat calls
The customer explains everything again to the next agent.
Engineer visits that were not needed
A visit is booked for a fault a line reset would clear.
Account issues disguised as faults
The service was suspended for non-payment.
Outage Check First, Then Only Your Approved Steps
The agent's order of work is fixed: known outages, then the account, then diagnostics from your runbooks. It runs a remote action only if it is on your permitted list and the customer agrees, and it reports honestly when the action did not fix the fault.
It does not change a customer's service, plan or equipment settings beyond that list, and account disputes go to the account team.
- Outages checked before diagnostics
- Steps from your runbooks only
- Permitted remote actions only
- Consent before any action
- Visits confirmed by your system
- One case per fault
The Systems Around It
The systems named here are the kind this agent works with. Which ones it connects to in your case is settled in the scope.
Network status
- Outage and maintenance data
Customer system
- Accounts, services and status
Network tools
- Line tests and permitted remote actions
Runbooks
- Your approved diagnostic procedures
Field service
- Engineer scheduling
Channels
- App
- web
- phone
Straight Answers
The pattern is shared: runbook, questions, one permitted fix, escalation. The helpdesk serves employees on company devices. This agent serves customers on your network, starts with an outage check and can end with an engineer at their door.
Utilities deal mainly with accounts, readings and outages. Telecom faults are diagnosed interactively with the customer's own equipment, which makes guided diagnostics the centre of this agent.
Where your network tools expose a test or a reset, and it is on your permitted list, yes, with the customer's consent. Where they do not, it works from what the customer reports and shows.
Not as software yet. This page describes a proposed workflow. It would reuse the runbook-driven diagnosis that runs in our IT helpdesk demonstration on test data: questions from an approved runbook, one permitted fix after consent, and escalation with a trace when it does not work. Line tests, outage data and engineer booking are not built. A project starts by mapping your process, then proving the first steps on test data before anything touches a live system.
Listen to Ten Fault Calls
Count how many minutes passed before the outage or account check, and how many steps the customer had already tried.
