Q&A: How AI Coaching Is Changing Customer Calls in Banking, Wealth, and Insurance

Banks, wealth managers, and insurers still hold many of their most valuable customer conversations by phone, and keep little of what's said. Quality teams typically score as little as 1% to 3% of contact center calls, according to ISG Research, so most conversations are never reviewed at all.

Naveen Gupta, Practice Director at Xebia, works with financial services firms on Xebia Customer Call Intelligence, a conversation intelligence solution that coaches representatives during the call and writes the outcome to the CRM. We asked him what gets lost when a call ends, how live AI coaching works without turning into a sales script, and where a team should start.

What is conversation intelligence, and why does it matter for banks, wealth managers, and insurers now?

Naveen Gupta: Conversation intelligence turns customer calls into structured data the business can act on: what the customer needs, how ready they are to decide, what was said about compliance, and what should happen next. It matters in financial services because the phone is still where customers bring their biggest decisions. They call about a loan, a renewal, money that's about to move. Until recently, the only way to learn from those calls was for a person to listen to a recording afterward, which almost nobody has time to do. Xebia Customer Call Intelligence does the work during the call, so the representative can use it while the customer is still on the line.

What actually gets lost when a customer call ends?

Naveen Gupta: Almost everything around the reason for the call. A customer phones to check a payment date and mentions, in passing, that their daughter starts college next year and another bank has quoted them a rate. The representative answers the payment question well. The college costs and the competitor quote stay in the recording, and the CRM records a disposition code. Multiply that by every call, every day. The need, the timing, the products the customer holds elsewhere, the objection they raised: none of it reaches the next person who speaks to them. That's why customers end up repeating themselves, and why follow-up treats every lead the same.

How does Xebia Customer Call Intelligence coach a representative during a call?

Naveen Gupta: Xebia Customer Call Intelligence transcribes the call as the customer speaks and turns what they say into fields: needs, timing, products held, life events, competitor quotes, objections. It places the customer on an intent ladder that updates as the conversation moves. Within a couple of seconds, the representative sees suggested responses, compliance prompts, and offers from the firm's own product catalog. Those suggestions keep updating as the customer talks. The representative decides what to say and whether to make an offer. Before the call ends, the next step is chosen, ranked against other open opportunities, and written to the CRM with a call brief for whoever follows up.

How fast is it, in practice?

Naveen Gupta: Words appear on screen within milliseconds of being spoken. Suggestions follow within a couple of seconds and keep updating as the conversation moves. That speed is the whole point. If a suggestion arrives after the customer has moved on to the next topic, it's a note for later, not coaching. The representative needs to see the right response while the customer is still talking about the thing that prompted it.

How does it help a representative make the right offer without pushing a sale?

Naveen Gupta: Three things keep it on the right side of that line. First, offers come only from the firm's own product catalog, so the AI isn't inventing products or terms. Second, the representative decides whether to make the offer. It's a suggestion they can ignore, not a quota. Third, the same pass that suggests an offer also checks the conversation for pressure selling, consent, and required disclosures. In financial services, an offer the customer didn't need is a compliance problem as well as a bad experience. The aim is the offer that matches what the customer just said, at the moment they said it.

How does it keep every representative on the same, compliant message?

Naveen Gupta: Every representative sees the same approved responses and disclosure prompts, whichever product the customer asks about. That closes the gap where two good representatives give two different answers to the same question. Then every call is checked for disclosures, consent, and pressure selling, not a sample. Compliance teams review the calls that were flagged instead of whichever calls a sample happened to pick. It's worth being precise here: Xebia Customer Call Intelligence checks and flags. It doesn't make a firm compliant on its own. The firm's compliance team still sets the rules and reviews what's flagged.

Can a person approve what the AI does before it changes records?

Naveen Gupta: Yes. Every automated step in Xebia Customer Call Intelligence, from coaching to writing records, can be set to require a person's review, and the firm decides which. A bank might let call notes go straight to the CRM but hold new leads for a loan officer to approve. Retrained models wait for a person's sign-off before they go live. Firms usually start with more review and loosen it as they see the results. I'd rather a client start cautious and relax than the other way around.

Does it replace the CRM or contact center platform?

Naveen Gupta: No. Xebia Customer Call Intelligence works on top of the tech stack a firm already runs: the CRM, core banking, loan origination, wealth and portfolio platforms, policy administration, and the product catalog. It reads the call and writes the results to those systems. It replaces none of them. That matters for adoption, too. Representatives keep working in the screens they know, and the data lands where the rest of the business already looks for it.

What did you learn from the consumer lending deployment?

Naveen Gupta: A US retail bank uses Xebia Customer Call Intelligence on its consumer lending calls. Before, loan details customers mentioned on calls, such as the amount, the purpose, and rates from other lenders, never reached the CRM. Every lead arrived in the callback queue with the same status, so follow-up was first-come, and quality assurance heard a sample after the fact. We connected live transcription and scoring to the bank's CRM, loan origination system, and callback queue. The biggest lesson was that the bank's lending and compliance teams had to own the thresholds, the offers, and the review steps. When they set those, representatives trusted the suggestions.

How would it work for a wealth management or insurance team?

Naveen Gupta: The same way, configured for each firm's products and rules. For a wealth team, Xebia Customer Call Intelligence listens for life events, goals, risk appetite, and money in motion. It can suggest a review meeting or a planning topic, and it checks suitability and best-interest disclosures, such as Reg BI. For an insurance team, it listens for coverage gaps, renewals, and life changes, can suggest added coverage, a renewal, or a bundled policy, and checks required disclosures and suitability for annuities. Consumer lending is where it's in use today. For wealth and insurance, we configure the fields, intent stages, offers, and checks with each firm.

What results should a firm expect, and how do you measure them?

Naveen Gupta: Independent research gives a useful starting point. A study of 5,179 customer support agents, published in The Quarterly Journal of Economics in 2025, found AI assistance raised productivity by 14% on average and 34% for newer agents. That study measured service work, not sales, so I wouldn't treat it as a forecast. Results depend on the team and the call flow, which is why we measure them in a pilot against a baseline agreed with the client. We compare offers made, conversion, time spent on notes, and compliance flags before and after, on the same call flow.

Where should a team start?

Naveen Gupta: With one call flow. Inbound loan inquiries, advisor review calls, or policy renewals are all good candidates, because they have a clear outcome and enough volume to measure. We configure Xebia Customer Call Intelligence to that team's products, offers, and review steps, run it, and compare the results with how the flow performs today. Starting small also lets the compliance team decide which steps need a person's review before anything scales.

Xebia Customer Call Intelligence Solution

Want to see what your calls are telling you? Book a demo of Xebia Customer Call Intelligence and walk through a live call, from the first word to the CRM record.

Sources