How to Build a Board-Level Business Case for an AI Voice Agent Programme in a UK Regulated Contact Centre

Arkadas Kilic 11 min read · 2026-06-28

Rel8 CX is an AWS Advanced Partner that builds autonomous AI voice agents for regulated UK contact centres, delivering production deployments in 4 to 6 weeks. We've sat in enough board rooms to know exactly where these business cases die, and this post is designed to make sure yours doesn't.

Most AI voice agent proposals fail at board level not because the technology is unproven, but because the case is built by someone who believes in the technology rather than someone who understands what the board actually needs to approve it. Finance wants a payback period. Risk wants a compliance answer. The CEO wants to know what happens when it goes wrong. If your deck doesn't answer all three in the first ten minutes, you've lost the room.

Here's how to build a case that gets approved.

Who Is Actually in That Room

Before you write a single slide, map your approvers and what each one needs to hear.

Your business case needs to speak to all four simultaneously. A deck that only addresses the financial return will get blocked by the CRO. A deck that only addresses compliance will get sent back by Finance for more numbers.

Start with the Problem, Not the Solution

This is where most proposals go wrong immediately. They open with "we want to deploy an AI voice agent" rather than "here is a documented operational problem that is costing us money and creating regulatory risk".

Your opening frame should be a current-state diagnosis. Pull real numbers from your contact centre:

• Average handling time (AHT): What's your current AHT for the top five call types by volume? In UK collections and financial services, we typically see AHT of 6 to 11 minutes for routine account queries that could be resolved in under 90 seconds.
• Containment rate on existing IVR: Most legacy IVR systems in UK regulated contact centres achieve 12 to 23% containment. The rest hits agents.
• Cost per contact: Industry benchmark for a UK contact centre agent fully loaded (salary, NI, benefits, training, attrition replacement, management overhead) runs between £8 and £14 per contact. Use your actual number.
• Volume by intent: Break down your inbound volume. In our experience across UK financial services deployments, 35 to 55% of inbound calls are routine: balance enquiries, payment confirmations, address updates, statement requests. These are automatable today.
• Complaint and re-contact rate: If customers are calling back because their query wasn't resolved, that's a Consumer Duty signal and a cost signal simultaneously.

Once you've laid out the current state with real numbers, the problem is self-evident. You're not selling AI, you're solving a documented operational problem.

The Financial Model: What the CFO Actually Wants to See

Don't present a single number. Present a model with three scenarios and show your assumptions explicitly.

Assumptions to State Clearly

• Containment rate target: For a well-configured AI voice agent on routine intents, a realistic first-90-days containment rate is 38 to 47%. We've seen 43% containment in week one on a UK debt management deployment. By month six, with tuning, that typically reaches 58 to 67% on the target intent set.
• Deflection value: Use your fully loaded cost per contact. Don't use a blended average if your deflected calls are cheaper to handle, use the cost of the specific intents you're automating.
• Volume: Use 12-month trailing inbound volume for the intent set you're targeting. Don't project growth unless you have a documented basis for it.
• Implementation cost: A production AI voice agent deployment on AWS with compliance controls, CRM integration, and a 4 to 6 week delivery timeline from Rel8 CX runs in the range of £65,000 to £120,000 depending on integration complexity. Ongoing infrastructure and support costs are typically £3,000 to £8,000 per month.

A Worked Example

Let's say your contact centre handles 180,000 inbound calls per year. 42% are routine intents suitable for automation. That's 75,600 calls. At £9.50 fully loaded cost per contact, that's £718,200 in annual handling cost for that intent set.

At 47% containment in year one, you deflect 35,532 contacts. Saving: £337,554.

At a deployment cost of £85,000 and £5,000 per month ongoing, year one total cost is £145,000. Net saving year one: £192,554. Payback: approximately 5.4 months.

Year two (assuming 62% containment from tuning): deflect 46,872 contacts. Saving: £445,284. Cost: £60,000 ongoing. Net saving: £385,284.

Three-year NPV at 8% discount rate: approximately £820,000.

Present this model in a table. Show the base case (conservative containment), the upside case (best-in-class containment), and the downside case (containment 20% below target). Even in the downside case, the payback period should be under 18 months if your intent selection is right.

The CFO will challenge your base case assumption. Have the data behind it. We can provide anonymised benchmarks from live UK deployments.

The Compliance Section: This Is Not Optional

In a UK regulated contact centre, the compliance section of your business case is not a supporting appendix. It's a primary gate. The FCA's Consumer Duty (effective July 2023 for open products, July 2024 for closed book) requires firms to demonstrate good outcomes for customers, including vulnerable customers. Any AI voice agent programme must address this directly.

What the Board Needs to See

If you can present a compliance section that answers all six of these points with specifics, your CRO will have nothing to block. Most proposals fail here because they treat compliance as an afterthought.

Addressing the Headcount Question Before It's Asked

Every board will ask: "What does this mean for our people?"

Don't dodge it. The honest answer in most deployments is that AI voice agents handle volume growth without additional headcount, and allow existing agents to focus on complex, high-value interactions. In some cases, there is a reduction in force. In others, the same headcount handles more volume as the business grows.

Be direct about which scenario applies to your organisation. Boards respect honesty. They don't respect evasion. If there is a redundancy programme attached to this, include the cost of that in your financial model (statutory redundancy, notice periods, HR support). It doesn't kill the business case. It makes it credible.

Also address agent experience. Agents who spend their day handling "what's my balance" calls are not engaged, and your attrition data probably shows it. Freeing agents to handle complex cases improves job satisfaction and reduces attrition cost. Attrition in UK contact centres runs 25 to 40% annually. At a replacement cost of £3,000 to £6,000 per agent, that's a real number worth including.

The Technology Section: Keep It Honest

Your board doesn't need to understand how large language models work. They need to understand three things:

1. What platform are we building on? AWS is the answer that lands best in UK enterprise and regulated environments. It's the platform most IT and risk teams already have a relationship with. Amazon Connect is a Tier 1 contact centre platform used by major UK financial services firms. Your board will have heard of AWS. They haven't heard of most AI voice agent vendors.

2. What does "production" mean? Be explicit that this is not a pilot or a proof of concept. It's a production deployment with enterprise-grade security, SLA-backed infrastructure, and compliance controls built in from day one. Pilots create scepticism. Production deployments create confidence.

3. What's the delivery timeline? 4 to 6 weeks from contract to production is the Rel8 CX delivery model. That's not a pilot timeline. That's a live deployment handling real customer calls. Boards are used to 12 to 18 month technology programmes. A 4 to 6 week delivery is a competitive advantage in the approval conversation.

The Risk Section: Show You've Thought About What Goes Wrong

Every board will ask about risk. Get ahead of it.

Showing this table tells the board you've done the risk work. Leaving it out tells them you haven't.

How to Structure the Board Deck

Keep it to 12 slides. Boards don't read appendices in the room.

1. The problem (current state, real numbers, cost and regulatory risk)

2. The opportunity (what automation makes possible)

3. The solution (what we're proposing, one paragraph)

4. Financial model (three-scenario table, payback period, 3-year NPV)

5. Compliance framework (Consumer Duty, data residency, audit trail)

6. Technology and delivery (AWS, 4 to 6 weeks, production not pilot)

7. Headcount impact (honest, specific)

8. Risk register (table format, mitigations for each)

9. Implementation timeline (weeks 1 to 6, what happens when)

10. Success metrics (containment rate, CSAT, AHT, complaint rate, cost per contact)

11. Governance (who owns this, how we report back to the board)

12. Decision requested (specific ask, budget, timeline, accountable executive)

Slide 12 is the most important. Be explicit about what you're asking the board to approve. "We are requesting approval for a £95,000 programme to deploy a production AI voice agent handling payment confirmation and balance enquiry calls, targeting 47% containment within 90 days, with a payback period of 5.4 months." That's a decision. Boards can approve decisions. They can't approve vague technology strategies.

Common Reasons These Business Cases Fail

What Happens After Approval

Once you have board approval, the clock starts. The board will expect to see results at the 90-day mark. That means your delivery partner needs to be capable of hitting production in 4 to 6 weeks, leaving you 6 to 8 weeks of live operation before your first review.

At Rel8 CX, our standard delivery model runs:

• Week 1 to 2: Discovery, intent mapping, compliance review, integration architecture
• Week 2 to 3: Build, CRM integration, Amazon Connect configuration
• Week 3 to 4: Testing, compliance sign-off, staging environment validation
• Week 4 to 6: Production deployment, live monitoring, first tuning cycle

By week six, you have a production AI voice agent handling real customer calls with full audit trail, vulnerable customer escalation, and CSAT measurement in place. That's what you take back to the board at 90 days.

The One Question Boards Always Ask Last

"What does good look like in 12 months?"

Have a specific answer. Not "we'll have improved our customer experience" but: "In 12 months, we'll be containing 62% of routine inbound volume autonomously, our cost per contact on automated intents will have fallen from £9.50 to under £2.10, our agents will be handling only complex and high-value interactions, and our Consumer Duty evidence pack will include 12 months of AI interaction data demonstrating good customer outcomes."

That's a vision the board can hold you to. And it's one you can deliver.

Frequently Asked Questions

Rel8 CX is an AWS Advanced Partner specialising in autonomous AI voice agents for regulated UK contact centres. We deliver production deployments in 4 to 6 weeks with FCA Consumer Duty compliance built in.

Rel8 CX delivers production AI voice agents on Amazon Connect in 4 to 6 weeks from contract signature. This includes CRM integration, compliance controls, and vulnerable customer escalation.

Based on live deployments, UK contact centres targeting routine inbound intents typically see 38 to 47% containment in the first 90 days, with payback periods of 4 to 8 months depending on inbound volume and cost per contact.

If you're building this business case right now and want a second opinion on your numbers, your compliance framework, or your board deck structure, we've done this before. We know what gets approved and what gets sent back.