Switchboard AI does not generate, negotiate, or commit to a customer-facing value or decision. Humans configure what is offered. Humans approve what is sent. Humans own every outcome. Built for FCA Consumer Duty, not retro-fitted to it.
How We Build It
Switchboard is tool-using AI in regulated work. These are the design constraints that make those tools safe.
Values, offers, and decisions are configured by named humans before the AI ever speaks. The engineer enters the total-loss valuation. The handler authors the email. The manager signs off the threshold. The AI presents what humans chose: it does not generate, decide, or negotiate.
When the AI shares comparable evidence (market valuations, similar cases, supporting data) it shares it both ways. Comps below the offer get shown. Comps above the offer get shown. The customer sees the full picture. The AI does not haggle, convince, or selectively withhold.
AI-drafted customer emails land in a handler approval queue before sending. pending_approval → approved → sent (or rejected). Never auto-sent. Always traceable to a named human who clicked approve.
When a caller asks for a person, or your rules say the call should go to one, it puts them through to your team. Anything it changed is already on the case.
Changes the agent makes are logged on the case as they happen. What your team does is logged against their names. Calls can be recorded, and the transcript emailed to the team.
The value the AI presents is one a named engineer set. The email it drafts goes out only when a named person approves it. The decisions that carry personal accountability stay with people.
Worked Example
The conversation pattern Switchboard runs in production. The AI presents the engineer’s value, captures the customer’s response, and routes disagreement back to the named human.
Before the AI calls the customer, a named engineer reviews the vehicle assessment and enters the total-loss valuation as a structured field on the case. This is the value the AI will present. The AI cannot change it.
The AI calls the customer, presents the engineer's value, and walks through the comparable market evidence symmetrically: comps below the value and comps above. The same evidence pack a human handler would use.
The agent writes the customer's answer to the case, and the change is logged. It doesn't push back, haggle or try to convince.
If the customer keeps disagreeing, it goes back to a person: the agent puts them through to the team, or the engineer rings them back. The AI never settles a dispute, and the recording and transcript are there to check.
In production at Laird Assessors. Laird Assessors runs its total-loss valuation conversations this way, alongside its out-of-hours calls and bodyshop chasing. The value the agent presents was set by an engineer, and disagreement goes back to a person.
The Hard Limits
Some come from how it's built: it can't send an email, and it can only use the tools it's given. The rest are rules we write into each agent and test.
No AI-generated total-loss valuations. No haggling. No convincing.
Indemnity, liability, repudiation, FOS-relevant calls: humans only.
Drafts go to the approval queue. A human clicks send.
It can only use the tools switched on for it. It has no other way into your systems.
Symmetric disclosure: comps both directions, every time.
Ask for a person and you're put through to the team.
Escalation
It's a straight transfer to your team, and anything it changed is already on the case.
“Can I speak to someone?” puts them straight through.
A complaint, distress, a dispute over a value: whatever you write.
Details that don't match, a case it can't find, a request it has no tool for.
Each agent puts calls through to the number you set for it.
Why This Design
Consumer Duty, SMCR, and audit defensibility are not bolt-ons. They shape what the AI can do, and, more importantly, what it cannot.
Fair customer outcomes, by design. Showing evidence both ways removes selective-information risk. Continued disagreement goes back to a person. The record builds up as the work is done, without a separate compliance project.
The decisions that matter are made by named people: the engineer who set the value, the person who approved the email, the manager who set the agent's rules.
The AI does not haggle, convince, or attempt to lower an offer. It presents and listens. Customer disagreement is a routing signal, not an objection to overcome.
Calls can be recorded and their transcripts emailed to the team. Changes are logged on the case, and each email approval is logged against the person who clicked it.
Buyer Questions
Because that's not what the AI is for. A total-loss valuation is a regulated decision with a real customer impact and a named human accountable for getting it right. The AI's job is to present the engineer's value with the same evidence pack a human handler would use, write the customer's response on the case, and put disagreement back in front of a person. Letting the AI generate or negotiate the value would put a regulatory decision in the hands of a system that cannot hold a SMCR responsibility.
The AI does not push back. It writes their position on the case and puts them through to the team, or the engineer rings them back. The recording and transcript are there for whoever picks it up. There is no haggling layer between the customer and the person accountable for the decision.
It follows the rules you write for it: what it's for, what it may do, and the calls it should put through to a person. That's the language model's judgement against your instructions, so we write and test the rules with you on real calls during the pilot.
Yes. The rules are plain English in Switchboard. Change them and the next call follows the new ones.
No. Switchboard is tool-using AI that makes real API calls during conversations: looking up cases, verifying identity, updating fields, creating new cases. The 'extra steps' are the design constraints that make those tools safe in regulated work: human-set values, evidence shown both ways, approval queues, a person on request, and a record on the case. A chatbot does none of that.
Switchboard runs in production at Laird Assessors: out-of-hours inbound calls, outbound bodyshop chasing, and total-loss valuation conversations. Its total-loss conversations follow the engineer-sets-value flow described on this page.
Pick one workflow: overflow FNOL, out-of-hours intake, or your bodyshop chase queue. Measured against your own SLA and cost-per-claim. No platform migration. No long lock-in.
