Contact centre AI business case template: what to include and what gets missed

AI business cases fail in predictable ways. Some count containment, the share of conversations that never reach an agent, and treat it as resolution, when a customer who gives up on a bot and calls back has cost more, not less. Some take a headline automation rate from a vendor or a conference and apply it across every contact, when the right figure differs journey by journey. Some count the AI's price and forget the knowledge, integration, testing and tuning that make it work. And some promise savings from agent hours without asking which hours, because the contacts left for people after automation are longer and harder.

A board does not need a perfect forecast. It needs a case that is honest about what is known, what is assumed and how the assumptions will be tested. The structure below follows the familiar five case model used for public sector business cases in the UK, adapted for contact centre AI, so that strategy, value, sourcing, affordability and delivery are each answered in turn.

Each section answers one question. Keep the main document short and put the journey-level working in an appendix, so a reader can follow the argument without wading through the model, and an analyst can check the model without rewriting the argument.

Strong benefits are measured from a baseline the organisation already trusts. Before any AI goes live, record for each journey the current volume, handle time, after-call work, transfer rate, repeat contact rate and customer satisfaction, from Genesys Cloud analytics or your existing platform. Then express each benefit as a change in one of those measures, with the assumption behind it stated plainly, and a range rather than a single figure where the evidence is thin.

Count resolution rather than containment: a contained conversation is only a benefit if the customer's need was met and they did not call back. For copilots, the clearest benefits are usually shorter after-call work, more consistent wrap-up, faster ramp-up for new agents and more interactions reviewed for quality. Some benefits are real but hard to value in money, such as availability outside opening hours or consistency of answers, and they belong in the case as described benefits rather than inflated into savings.

• Resolved contacts, confirmed by low repeat contact, not just contained ones
• Reduced after-call work and handle time where copilots draft summaries and find knowledge
• Faster ramp-up for new agents and more consistent answers
• Wider quality coverage from AI scoring, with people coaching
• Service outside opening hours for journeys that suit it

The AI's own price is the easiest cost to find and rarely the largest over the life of the service. On Genesys Cloud CX, native AI is metered in AI Experience tokens, and every organisation's free allowance is shared by all AI features, so per-user features such as Agent Copilot can use it up before a single bot interaction is counted. Third-party voice AI carries a Genesys per-minute platform charge on top of the vendor's own price. Our AI Experience tokens guide explains the model, and our voice AI cost calculator compares routes per resolved call at published list prices.

Then there is the work around the AI. Knowledge must be written and kept current, because AI answers are only as good as the content behind them. Integrations must be built so that AI agents can act safely. Testing takes longer than for scripted flows, and must be repeated when models, prompts or knowledge change. Someone must review AI conversations, tune the design and own the guardrails after go-live. These are recurring costs, not project costs, and a case that leaves them out will look better in year one and worse every year after.

1. Tokens beyond the shared free allowance, including per-user features
2. Platform charges for third-party AI, on top of the vendor's own fees
3. Design, integration and data action build
4. Knowledge creation and upkeep
5. Testing at launch and after every significant change
6. Ongoing review, tuning and guardrail ownership
7. Second-vendor contract, support and data protection work, where applicable

Boards now ask about AI risk as readily as about cost. The case should name the main risks for each journey and the control for each: wrong or invented answers, controlled by grounding in governed knowledge, guardrails and testing; unsafe actions, controlled by narrow tools and permissions; data protection, controlled by a data protection impact assessment and agreed processing terms; and service failure, controlled by a fallback to people. Regulation belongs here too: the EU AI Act's transparency duties for AI that talks to people now apply, and UK GDPR rules on automated decisions apply where a decision has a legal or similarly significant effect. Our guide to the EU AI Act and UK GDPR for contact centres covers what applies and when.

Vendor and model dependency deserves its own line. AI models change: Genesys versions the agent model behind agentic virtual agents and offers an opt-in period before a new version becomes the default, and third-party providers change models and prices on their own timetables. The case should say who will test each change before it reaches customers, and how the organisation could move a journey to another route if it had to.

The most persuasive business cases ask for the right to prove the next step rather than funding for everything at once. Choose a first journey with enough volume to measure, a clear outcome and limited risk. Agree success criteria and the measures that will decide them before the pilot starts, and set a decision point at which the evidence is reviewed and the next phase approved, adjusted or stopped. Then update the case with real figures, which turns the next request from a forecast into a track record.

QVCCS is a Genesys referral partner, not a reseller, and we take no commission from any AI vendor, so our part is the evidence rather than the sale. We help clients build the journey model behind the case, test candidate routes with real calls placed from outside the platform, and set up the measures that show whether the benefits arrived. Our questions for your Genesys Cloud business case guide lists what each stakeholder is likely to ask.