Business case forAI in customer service
The math finance wants to see: what customer service costs today, what it will cost with AI, how soon the investment pays back and what could go wrong.
Worked example
Today
$12,500
With AI
$8,040
Savings per month
$4,460
Payback
less than 1 month
Business case is the document that justifies an investment with cost, return and risks. For AI in customer service, it fits on one page.
Reviewed on 11/16/2026
Four blocks, in this order
01
Current cost
How much each conversation costs today with people. Use salary plus taxes and benefits and, for a fuller picture, add tools and management.
Cost per conversation
= monthly cost per agent ÷ conversations per agent
Current cost
= conversations per month × cost per conversation
02
Cost with AI
AI resolves part of the conversations and the rest stays with the team. Add both parts.
AI conversations
= conversations per month × % resolved by AI
Cost with AI
= human conversations × cost per conversation + platform cost
Monthly savings
= current cost − cost with AI
03
Payback and return
How long the savings take to pay back the initial investment, which is mostly your team's time to build the knowledge base, test and integrate.
Payback (months)
= initial investment ÷ monthly savings
12-month return
= (annual savings − initial investment) ÷ initial investment
04
Risks and mitigation
The CFO approves faster when each risk already comes with its answer.
The AI gives wrong answers
A reviewed knowledge base, simulations before publishing, a trail for every answer and rule-based handoff to people.
AI resolution falls below forecast
A pilot with a resolution target and a review date before expanding.
Cost grows with volume
Simulate two or three volumes and check the overage price in the proposal.
Security and LGPD requirements
Security questionnaire answered before signing.
Low adoption by the team
Training, defined roles and the AI taking over repetitive topics first.
An operation with 3,000 conversations per month
Illustrative numbers, using the same math and public prices as the ROI calculator. Replace them with yours.
Assumptions
- Conversations per month
- 3,000
- Monthly cost per agent (salary, taxes and benefits)
- $2,500
- Conversations per agent per month
- 600
- Resolved by AI
- 50%
- Initial investment (120 team hours)
- $1,705
Result
- Cost per human conversation
- $4.17
- Current cost per month
- $12,500
- Conversations resolved by AI
- 1,500
- Tolky platform (Starter plan)
- $540
- Cost with AI per month
- $8,040
- Monthly savings
- $4,460
- Payback
- less than 1 month
- 12-month return
- 3,040%
On average, AI resolves more than 67% of interactions across the Tolky platform. The example uses 50% to stay on the conservative side.
Purchase memo for the CFO
A one-page memo. Copy it, fill in the fields in brackets and send it.
Subject: Proposal for an AI agent in customer service To: [CFO's name] From: [your name and department] Date: [date] 1. Problem Today we handle [volume] contacts per month with [number] people. Each conversation costs [amount], and [main pain point: response time, after-hours queue, lost sales]. 2. Proposal Hire [vendor] so an AI agent resolves repetitive conversations on [channels] and hands off to the team, with the full history, whatever requires a person. 3. Cost and return Current cost: [amount] per month. Cost with AI: [amount] per month, of which [amount] is the platform. Estimated savings: [amount] per month and [amount] over 12 months. Initial investment: [amount], with payback in [months]. Assumption: AI resolves [%] of conversations. 4. Risks and mitigation Answer quality: tests before publishing and rule-based handoff to people. Security and LGPD: security questionnaire answered on [date]. Resolution below forecast: [duration] pilot with a [%] target and review on [date]. 5. Decision requested Approve the [duration] pilot, with a budget of [amount], and reassess on [date] with the results.
Questions about the business case
Compare the current cost (conversations × cost per human conversation) with the cost with AI (conversations that stay with people × cost per conversation, plus the platform cost). The difference is the monthly savings. The initial investment divided by it gives the payback in months.
A conservative number, validated later in the pilot. At Tolky, AI resolves more than 67% of interactions on average across the platform, but the result depends on the type of service and the knowledge base. The example on this page uses 50%.
Mainly your team's hours to organize the knowledge base, test the agent, integrate systems and train the people who will operate it, plus implementation services, if hired.
Tolky charges per conversation and has a free plan, with 150 conversations. Starter begins at $90 per month, with 500 conversations, and Professional at $450 per month, with 1,000 conversations. Enterprise is custom.
The other pieces of the kit
For the evaluation team
RFP template
11 sections and 40 questions to compare any AI agent vendor on WhatsApp, with what a good answer includes.
For security, IT and DPO
Security questionnaire
Tolky's answers, including what it does not have today, and the questions worth asking any vendor.
To close the math
ROI calculator
Enter your volume and team cost and see the estimated savings with Tolky's public prices.
Pages that answer the rest
Do the numbers add up?Let's do it together
In the demo, we build the business case with your volume, your team's cost and the right plan.