8 recommendations for selling Conversational AI to banks

Hello dear reader.

This is a post I published on Conversational AI News a few days ago. If you'd like to make sure you get updates immediately, please sign up (free) for an account on the site.

Right, let us begin!

I am working with multiple banks helping them with their Conversational AI strategies at the moment.

Every firm I'm supporting is frustrated by the vendors they've been interacting with, to some degree.

They're not getting answers to their questions online. Many are carefully avoiding giving their contact details to avoid the "automated-hell" sales processes.

Those who have had some exploratory vendor discussions have found the process unsatisfactory.

To this end, here are some high-level recommendations for vendors seeking to optimise their approach to selling to the financial services marketplace. This is non-exhaustive and is based on what I'm seeing anecdotally through multiple conversations with institutions in different regions.

(Note: I'm using banks as a catchall term for FinTechs-with-banking-licenses, Banks, Insurance companies, Brokers and Asset Managers and so on.)

First, some context: As a former C-Level bank digital/technology/transformation executive, I have direct experience as the ultimate decision maker or the #1 key influencer. Now, I regularly advise clients working with Conversational AI: I offer advice, independent assessments, a comprehensive maturity framework, vendor beauty parades and full support through the whole purchase process from RFI to production live.

Based on recent discussions I've been having, here are some areas where vendors might consider focusing their attention to better attract and win in the financial services marketplace:

1) Understand The Client's Motivation

I'm regularly surprised by the groupthink happening inside Conversational AI vendors. Whether it's marketing, sales, product or the vendor's executive team making assumptions, there is often little awareness or experience of how financial services operate or think when it comes to Conversational AI.

There's a lot of nuance of course, but the answer is simple for banks: It's the bottom line.

• How many FTE can I re-deploy or save, as a result?
• Will it cut my customer queues or waiting times?
• How good is the tech? Says who?
• How do I integrate into my wider service offering?
• How workable is it in my jurisdiction?
• What does your roadmap look like? (see my roadmap point below)

Recommendation: Hire some former financial services executives as advisers to help improve and challenge your approach.

I regularly offer this service to vendors looking for perspective. Ask the adviser to do a mix of the following:

• Get them to look through your material and challenge it
• Get them to explain why your standard automated approach to developing your sales and marketing pipeline isn't working too well
• Get them to listen to your practice sales pitches and offer suggestions for improvement
• Ask them to sit and work with your product management and technology teams to explain how financial services generally work in terms of priority and approach
• Ask them to work with your marketing and communications teams to explain how they tend to use websites, white papers, webinars, events and so on
• Ask them to explain how they and their colleagues like to be engaged by vendors

Example: Here's a sample quote from a European banking executive I've been working with recently:

I just don't know where to start: All the [vendor] websites are talking about Generative AI and other terms that we just can't support. I just want the best chatbot for my bank. What vendor should I talk with?

2) Answer The Cost Question Quickly

Almost every vendor has a problem with this one. You know it's going to be annoying when you click on "Pricing" and the page says "Please Call".

That's understandable. Pricing is one thing, though. Business models... that is where you might give some thought to make things even better.

When you're dealing with financial services clients, your Conversational AI business model is (usually) infuriating. I know you like to think it's great. You think your 'shared value' or 'shared risk' business model works brilliantly. It does: For you, and your board of directors. One of the reasons you are constantly fire-fighting with Financial Services clients is that their business and funding approach doesn't work in the same way.

How much does it cost? Give me an annual figure or a fixed price.

And if you think you've pulled the wool over my eyes and managed to get me to agree to your crazy billing approach for 3 years, that's just because I didn't prioritise a complex, heavy negotiation this time. I will swallow the issue short-term and then absolutely nail you to the wall in month 20 of our 36-month agreement. Just wait. I am seeing this happen often with vendors. Everything feels good - you think there's no way the bank will leave. Then you get the phone call. The only vendors having an OK time at renewals right now are those that agreed on sensible, fixed(ish) terms.

Recommendation: Have an answer to the 'How much is it for unlimited?' question. The major issue with per-usage models in Conversational AI in Financial Services is that (in general) budgets in banks don't work the way you think they do - and that's not changing soon. Sit with your industry advisers and get their perspectives. Think back to every painful negotiation you've had with an FS client. Compare notes. Figure out a better, clear approach to pricing.

3) Don't Rely On Analyst Report Status

The vendor marketing director will be delighted to see their brand named in the top right box by an analyst firm. It's deeply alarming when you're in preliminary sales discussions with a vendor and the only thing they keep repeating are the various ratings from industry analysts who have:

• Never, ever deployed Conversational AI in production
• Aren't staking their personal reputation, bonus or career on your product
• Don't have millions of customers ready to complain to regulators and the media the moment there's a public screwup
• Don't have a bottom line to protect

So yes, in the financial services industry, having the right badge(s) is nice. It's good. One might suggest it's even table stakes. But it doesn't get you as far as you think it does.

Recommendation: Prioritise real-life case studies from other financial services clients if you have them, failing that, give examples from your high-profile, high-volume clients. Answer the burning questions on your website – don't save it until the sales calls. Many of the banks I'm working with are making their high-level qualifying decisions based on your websites and what they're hearing from friends and colleagues.

4) Evidence A Stable Product Roadmap

Do you have a stable product roadmap? Is it published? Can we have a look? Or are you going to change it based on the whims of whoever talked to your Chief Product Officer or Head of Product recently?

Yes, it's AI, yes, the AI marketplace is changing fast. But that doesn't mean you need to run with every single fad that OpenAI or other AI leaders have thrown into a press release to justify their latest multi-billion raise.

Too often, my banking clients are alarmed by frequently changing product roadmaps (or definitions!). One minute it's a platform, the next minute it's an API, then it's a Service Offering Via Partners. Banks are looking for certain, predictable and stable. They'll want to know you're aiming to lead the field with future functionality but equally, they will also want the ability to opt out of some features too.

Recommendation: Publish a public roadmap - or if that's too radical, publish a private roadmap that you share on an NDA basis.

5) Deal With The Hallucination Issue

This is an issue for some banks.

Some I speak to don't allow the word hallucination to be used internally when they're discussing AI. It's that dangerous.

This is because using the word can underplay the critical challenge that all regulated institutions face: Hallucination means incorrect.

For the past few decades, no one has had to worry whether software would start giving incorrect answers. It just wasn't a problem.

And now there are committees sprouting up everywhere across the small and the huge, all focused on AI model review and the removal of any hint or prospect of anything going into production that can't add up.

In one memorable I saw one bank shelf their super-awesome Generative AI proof-of-concept because the LLM kept on getting confused between two functions: SUM and AVG. The last thing you need is your LLM arbitrarily changing the interest calculation.

So there is already a huge amount of learned wisdom embedded in financial services when it comes to Generative AI. Yes, it's fine when it's summarising an email, or making internal meeting notes. But it is not fine to give incorrect answers when talking to customers. This plays directly in all interactions they will be having with you.

Be incredibly clear about how and when your system is using Generative AI to give answers to customers. Get ready to handle the liability question – if your system gives our customers the wrong answer, will you indemnify us from the financial losses? That's when the negotiation can get interesting.

Recommendation: Clearly explain how your systems answer customer questions – and how those answers are programmed. Banks need to be able to relay this to their senior executives, internal risk, compliance and audit teams – and often to the industry regulator.

6) Don't Expect Customers To Be Call Centre Experts

I recently had one executive I was working with stop me and say, "What does containment mean?"

This individual was supervising their bank's Conversational AI strategy.

They haven't worked in call centers or contact centres. They don't know the lingo.

Give strong thought to the language and terminology you're using online, in videos, webinars and in any other materials.

I am seeing a lot of interest in Conversational AI from individuals who don't identify, in any way, with anything to do with call centres. This is especially important to consider for those vendors that evolved their offerings from the call centre domain. Sometimes the product offering and associated materials are almost impenetrable to those from outside the domain.

Recommendation: Make sure you use simple, accessible language that will make sense to financial services executives who don't have a call centre background.

7) Don't Forget Internal Deployments

A lot of the players I'm working with are choosing to evaluate Conversational AI for their internal teams first for many reasons, mostly around risk and confidence with the technology.

Some of the best successes I can mention with Conversational AI are from internal implementations (e.g. Agent Assist).

Recommendation: Make sure your marketing material and sales approach is geared to support and encourage internal usage, not just external customer-facing services.

8) You And Who's Army? What Does It Take To Go Live?

Be clear with banks about how to implement your technology.

What does the golden (ideal) path look like?

What are the likely gotchas that I'm not thinking about right now? (e.g. Access to core banking systems).

Quite often I feel like writing a how-to guide for the banking industry for Conversational AI vendors. I haven't been able to dedicate the time yet. But here's what I like to do: I'd like to sit down and look at the typical vendor deployment approach and then put this in terms that financial services players can rapidly understand. For example, clients in other industries can often just press a few buttons and give their vendors access to "the CRM system". Most banks don't have one single source of truth for their customers. Sometimes it's product-dependent (e.g. Savings, Credit Card) and other times the data is arrayed across multiple core systems. This is important to consider when you want the chatbot to answer questions about the customer's products.

Your financial services adviser(s) can help you construct this kind of golden path approach.

Do I just get the license from you? Or are you going to fly in a team of YOUR employees to help me? Or do you have a roster of preferred implementation partners? Who are they? How good at they?

Recommendation: Do the heavy lifting for customers now – show them the best way of deploying your technology in a financial services environment. If your software is best deployed with a partner, make that clear upfront.

9) Bonus - Some Other Points To Consider

Here are some other areas where I see vendors coming unstuck:

• Misunderstanding the importance of existing regulations such as GDPR, EU AI Act, etc.
• Making blanket statements about compliance without addressing jurisdiction-specific needs - I need to quickly ascertain if I can use your tech in my region
• Failing to address how systems handle sensitive financial data across different regulatory frameworks - most regions have very specific rules and processes for this, so try and address how your systems work
• Cybersecurity issues - everything from security certifications, standards and protocols, including limited or no recent information on penetration test reports, external security audits or code reviews
• Related to this: Insufficient detail about audit trail, logging capabilities and internal admin/user controls
• Help with end-user and employee training and the related technology adoption/change management processes

Thanks for reading. If you're working in a financial services company and you'd like some advice, please reach out - there's no cost for an initial consultation. Likewise, if you're representing a vendor - please get in touch for a chat .

Ewan