Don't Buy AI Support on Vibes

"AI will save you money" is true in the same way "exercise is good for you" is true. Technically correct but useless without specifics. Before committing to any AI support tool, run the actual numbers for your business.

Here's the framework.

What You're Spending Now

Figure out your current cost per support interaction. If you have dedicated support staff, the math is:

(Annual salary + benefits + tools) / (tickets handled per year) = cost per ticket

A support agent making $45,000/year with $10,000 in benefits and $2,000 in tool costs handles roughly 3,000-4,000 tickets/year. That's $14.25-$19 per ticket.

If you (the founder) are doing support yourself, your cost is higher because your time is worth more. What else could you be doing with those 2 hours a day? If your time is worth $100/hour and you spend 10 hours/week on support, that's $4,333/month in opportunity cost.

If you're using outsourced support, the cost is whatever you're paying per hour or per ticket. Typically $8-25/hour or $6-15/ticket.

What AI Support Costs

Most AI support tools charge one of two ways:

Per resolution: Intercom Fin ($0.99), Zendesk ($1.50 committed), Gorgias ($0.90 annual), Supp ($0.30). Plus any base fees or seat costs.

Per seat with AI included: Help Scout ($50/month contact-based, unlimited users), Freshdesk ($49-79/agent/month for Pro/Enterprise, plus $29/agent for Freddy Copilot).

For per-resolution tools, your cost is: (monthly tickets x expected automation rate x cost per resolution) + base fees + seat fees.

The Expected Automation Rate

This is where most ROI calculations go wrong. Vendors claim 70-80% automation rates. Reality for most teams in the first few months is 20-40%.

Why the gap: vendors measure "resolution" generously (the AI responded and the customer didn't come back, but maybe they just gave up). And their stats come from mature deployments with well-trained AI, not fresh installations.

Use these estimates for your first 90 days:

Simple support (e-commerce, order status, FAQ): 35-50% automation rate.

Medium complexity (SaaS, billing + technical): 20-35% automation rate.

High complexity (B2B, enterprise, regulated): 10-20% automation rate.

Service businesses (salons, restaurants, repair): 40-55% automation rate. Mostly repetitive booking and hours questions.

After 6 months with tuning, add 10-15 percentage points.

Running the Numbers

Let's work through an example.

A startup with one founder doing support. 400 tickets/month. Current cost: 15 hours/week at $100/hour opportunity cost = $6,500/month.

Scenario A: Intercom Fin. Essential plan at $29/seat/month (just the founder). 400 tickets, 35% automation rate = 140 AI resolutions at $0.99 = $138.60. Plus $29 seat fee. Total: $167.60/month. Time saved: 35% of 15 hours = 5.25 hours/week.

Scenario B: Supp. No base fee. 400 tickets, 35% automation rate = 140 resolutions at $0.30 = $42. Plus 260 classifications at $0.20 = $52. Total: $94/month. Same time saved: 5.25 hours/week.

Monthly savings with Scenario A: $6,500 - $167.60 - (9.75 hours x $100 x 4.33 weeks) = roughly $2,100/month net savings (reclaiming 5.25 hours/week worth $2,275/month, minus tool cost).

Monthly savings with Scenario B: $6,500 - $94 - same time math = roughly $2,181/month.

Both options save roughly $2,100-2,200/month in this scenario. The tool cost is minor compared to the time savings.

When AI Support Doesn't Save Money

If you get fewer than 50 tickets/month, the time savings are minimal. You might save 1-2 hours a week. Worth it if you hate doing support, but the ROI calculation won't be dramatic.

If your tickets are almost all complex (requiring judgment, investigation, or multiple follow-ups), AI won't automate much. A B2B company where every ticket requires looking up a customer's specific integration and debugging their API calls won't see 35% automation.

If your team is already fast and efficient, the baseline cost per ticket is lower, so the savings are smaller.

If your support volume is seasonal (holiday spikes but quiet otherwise), the monthly cost of AI tools during slow periods may not justify the savings during busy ones. Run the math on your slowest month, not just your busiest.

The Quick Formula

Monthly savings = (tickets x automation rate x current cost per ticket) - (tickets x AI cost per ticket) - base fees

If that number is positive, AI support saves you money. If it's more than 3x the tool cost, it's a clear win. If it's close to break-even, the value comes from speed (customers get instant responses) rather than cost savings.

Run it with your numbers. If the math works, try the cheapest option first for 30 days and see if the real automation rate matches your estimate.