AI chatbot vs human support: the real cost math for a 100-customer SaaS

By Saurav | Founder of saavos | Building in public toward $10k MRR

[!TLDR] At 100 customers generating 300 support interactions per month, a part-time VA costs $640–2,000/month. An AI chatbot costs $19–40/month and deflects 40–60% of that volume. The math isn't "chatbot OR human" — it's "chatbot first, then human for what the bot can't do." Deploy the chatbot to get deflection-rate data before you hire.

What "100 customers" actually costs in support

A SaaS product at 100 customers with a modest support load gets roughly 2–5 support interactions per customer per month in the early stage. Call it 300 interactions/month — a middle estimate, consistent with Zendesk's benchmark reports for SMB SaaS. Some are quick ("how do I change my plan?"), some take 20 minutes (debugging a weird edge case).

If you handle those yourself as a solo founder, you're spending 30–60 hours a month on support. At any reasonable opportunity cost — even $50/hour, which undersells a founder's time — that's $1,500–$3,000/month of founder time gone.

If you hire:

• A part-time VA doing email support: $15–25/hour x 20 hours/week = $1,200–2,000/month
• A dedicated part-time support contractor (US-based, async): $25–35/hour x 15 hours/week = $1,500–2,100/month
• Offshore async support (Philippines, India, experienced): $8–15/hour x 20 hours/week = $640–1,200/month

Those are real market rates as of 2026 — industry standard, your mileage varies by platform and contract type. The numbers move around but the ballpark holds.

Now here's the chatbot side.

What a chatbot actually costs at 100 customers

A RAG-grounded chatbot — one trained on your actual docs, not just a generic LLM — deflects a specific category of questions. The ones it deflects are the repeat questions, the ones your docs already answer, the FAQ hits. Industry estimates put 40–60% of inbound support volume at early-stage SaaS in this bucket.

So at 300 interactions/month, a chatbot realistically handles 120–180 of them. The remaining 120–180 are the hard stuff: bugs, billing disputes, edge-case configuration, angry users, anything the docs don't cover. Those still need you or a human.

The cost of the chatbot:

• saavos Starter plan: $19/month flat. Handles 1,000 messages/month, so 300 interactions isn't even close to the limit.
• Chatbase Hobby: $40/month.
• SiteGPT Starter: $39/month. 4,000 messages — same story.
• FastBots.ai Essential: $16–19/month.

So the chatbot slot is $19–40/month. That's not a typo. The monthly cost of the software is roughly 1–2 hours of a human support person's time.

The actual comparison

Here's what I'd lay out if I were making this decision:

The most interesting row is the last two. A chatbot doesn't replace a human support person — it changes what the human spends their time on. If you have a VA, a chatbot means the VA is only handling the hard stuff. You're paying the VA less per resolved ticket. You might need 10 hours instead of 20.

That's where the math gets interesting. Not "chatbot OR human" — but "chatbot first, then human for what the bot can't do."

What the chatbot doesn't do

I want to be specific about this, because a lot of chatbot marketing fudges it.

A chatbot grounded in your docs handles: feature explanations, pricing questions, how-to walkthroughs that mirror your documentation, common error messages that have documented solutions, setup instructions.

It doesn't handle: billing disputes, refund requests, emotionally charged customers, multi-step debugging that requires back-and-forth, anything that requires account access or live action.

Those require a human. Period. No chatbot — mine or anyone else's — is getting you out of that.

So if your 300 monthly interactions are 60% FAQ-category and 40% complex or emotional or account-specific, the chatbot handles 180 interactions and you still need a human for 120. The chatbot just reduced your human-handled volume by 60%, which either frees founder time or reduces the VA hours you need to buy.

Why this math matters at the indie SaaS stage

At 100 customers you're probably not ready to commit to a full-time support hire. The floor for meaningful part-time help is $600–800/month. That's real money when you're pre-revenue or sub-$1k MRR.

The chatbot at $19/month is the thing you deploy before you need the hire, to figure out whether you actually need the hire. You might find that 60% deflection means your founder support time drops from 40 hours/month to 16 — and at 16 hours, you handle it yourself until you hit 300 customers.

Or you deploy the chatbot, realize your customers have a lot of complex billing questions (not deflectable), and decide to hire earlier.

Either way, $19/month to get that signal is cheap. The alternative — hiring before you have the deflection data — is how founders over-hire.

One thing I got wrong

I assumed the setup cost would matter. It doesn't, really. A RAG-trained chatbot pointed at your docs takes about 5 minutes to configure and maybe an hour to tune the initial prompts. That's not a meaningful factor in the monthly cost comparison.

The meaningful variable is deflection rate — which you can only measure after you deploy. If your bot deflects 20% instead of 60%, the math changes. If it deflects 80%, you probably don't need the VA hire this quarter.

Deploy the bot first. Measure. Then decide on the hire.

---