Replace Drift Pricing and Intercom for Under $100/mo

TL;DR

Intercom Pro costs $74/seat/month. Drift pricing starts at $2,500/year for its Essentials plan and climbs quickly into five-figure territory for anything beyond the basics. A Claude API chatbot handling 10,000 messages costs $15-40/month total. If your chatbot use case is answering common questions and qualifying leads, the custom build pays for itself inside 60 days.

What Intercom Actually Costs You

Intercom Pro is $74/seat/month. If you have 2 people handling chat, that is $148/month, or $1,776/year, before you hit any message volume limits or add-ons. The per-seat model scales against you as your team grows, and Intercom has a track record of annual price increases that catch customers off guard at renewal.

For most SMBs, the actual usage pattern looks like this: 80% of chat volume is the same 15 questions, 15% is lead qualification, and 5% is stuff that genuinely needs a human. You are paying enterprise pricing to automate a problem that does not need enterprise software.

Where the Hidden Costs Appear

Intercom charges for message volume overages once you exceed your plan tier. Proactive messaging, product tours, and their AI Fin feature all sit behind additional line items. A business that signs up expecting the base $74/seat price routinely ends up paying 30 to 50 percent more once those add-ons get switched on during onboarding. If you are running any kind of outbound chat sequence, the costs compound further.

What SMBs Are Actually Using

Most SMB teams using Intercom are relying on three things: the live chat widget, the inbox for routing conversations to the right person, and the automated bot responses for after-hours coverage. The CRM features, customer data platform, and product analytics are collecting dust in the majority of small-business accounts. That is a lot of money for a chat widget with some routing logic.

What Drift Pricing Looks Like in Practice

Drift is more opaque than Intercom, and that opacity is deliberate. The company pushes prospects toward sales calls before revealing numbers, which is itself a signal about where the pricing lands.

Drift Pricing Tiers Broken Down

Drift’s Essentials plan starts at approximately $2,500/year billed annually, which works out to around $208/month. That covers basic live chat and one seat. The Premium tier, which unlocks AI features and more robust routing, is only available via a sales quote and typically starts around $12,000/year for small teams. The Advanced tier runs $40,000 or more annually for larger organizations.

For the SMB owner comparing tools, the practical answer is this: Drift pricing for anything beyond the entry-level plan is more expensive than Intercom, and the entry-level plan is deliberately limited to push upgrades. If a vendor will not show you pricing without a demo call, the pricing is not competitive.

What You Actually Get From Drift

Drift’s core value proposition is conversational marketing, meaning it is designed to qualify and route sales leads, not handle support volume. If your chat is primarily support-focused, you are buying the wrong tool regardless of price. If your chat is sales-focused, Drift does that reasonably well, but the cost-per-qualified-lead math still needs to be run against what a custom Claude-powered qualification flow would cost.

The Claude Alternative: What It Costs

Claude 3 Haiku is Anthropic’s fastest, cheapest model. At 10,000 messages per month with average message lengths, you are spending $15-40/month in API costs. That is not a typo.

The Annual Cost Comparison

Intercom Pro at 2 seats runs $1,776/year. Drift Essentials for 1 seat runs $2,500/year. A Claude chatbot at $30/month average runs $360/year. The delta between a Claude build and a 2-seat Intercom account is $1,416 annually. The delta against Drift Essentials is $2,140 annually. You own the Claude build outright, which means no renewal negotiation, no seat-count surprises, and no vendor lock-in.

Build cost runs 8 to 15 hours if you use n8n for the backend and a simple widget for the front end. At any reasonable contractor rate, you break even in under 3 months.

API Cost Scaling Explained

The $15-40/month estimate assumes Claude 3 Haiku at roughly 500 input tokens and 300 output tokens per exchange, which is typical for a support or lead-qualification conversation. If your volume is 50,000 messages/month instead of 10,000, the cost scales to approximately $75-200/month. Even at that volume, you are still below a single Intercom seat, and well below Drift’s entry-level annual plan.

The Stack You Actually Need

You do not need a full-time developer for this. Here is what the build looks like.

Front-End Widget Options

The front end is a chat widget. You can use Tidio’s free plan as a shell and swap the AI brain behind it, or you can embed a simple custom widget on your site in about 2 hours. The widget just needs to send a message and display a response. For most SMBs, a lightweight JavaScript widget that opens a chat panel, captures the user’s message, sends it to a backend endpoint, and renders the response is the entire front-end build.

Backend Architecture With n8n

The backend is n8n, either cloud or self-hosted. A webhook receives the message, sends it to Claude’s API with your system prompt and context, and returns the response. The whole workflow fits on one n8n canvas. n8n Cloud starts at $20/month, and the self-hosted version is free if you have a server already running.

For conversation memory, you can store a session array in n8n’s static data or write each exchange to a simple Supabase table. Retrieving the last 6 to 10 messages and passing them back to the API on each request is enough to give the bot short-term conversational context without blowing up your token costs.

System Prompt Engineering

The system prompt is your product. This is where you feed Claude your FAQ content, your escalation rules, your tone, and your business context. A well-written system prompt is the difference between a bot that sounds like your company and one that sounds like a generic AI.

Start with a clear role definition: tell the model what it is, who it serves, and what it is not allowed to do. Then add your FAQ content in clean, direct prose. Avoid pasting in raw documentation or policy PDFs. Summarize the key points in the same tone you would use in a conversation. Finish with explicit escalation instructions: if the user asks about billing disputes, account cancellation, or anything outside the defined scope, the bot should collect a name and email and route to a human.

Knowledge Base Structure

The knowledge base can be a Google Doc, a Notion page, or a Supabase table. For most SMBs, a detailed text file injected into the system prompt handles 90% of cases without any retrieval complexity. If your knowledge base exceeds roughly 4,000 words, consider a simple retrieval layer using Supabase vector search, which adds about 3 to 4 hours to the build but keeps token costs low at scale.

What You Give Up

Honest answer: you lose Intercom’s built-in CRM, their shared inbox UI, and their customer data platform. Those are real things.

But audit whether you are actually using them. Most SMB teams using Intercom are using the chat widget and routing tickets to email. The CRM features collect data nobody looks at. If that is you, you are paying $1,776/year for a chat widget with AI that you can replicate for $360.

If you are running a 10-person support team with complex routing, SLAs, and Salesforce sync, do not build this. Intercom earns its price at that scale. Same applies to Drift: if your sales team is running multi-threaded account-based plays with deep Salesforce integration and you have budget to match, Drift’s Premium tier may be worth the cost. For SMBs with under 5 support or sales agents and under 20,000 messages per month, it almost certainly is not.

How the Costs Stack Up

Option	Monthly Cost	Annual Cost	Setup Cost	Build Time	Control
Intercom Pro (2 seats)	$148	$1,776	$0	2-4 hrs onboarding	None
Drift Essentials (1 seat)	$208	$2,500	$0	3-5 hrs onboarding	None
Drift Premium (2 seats, est.)	$1,000+	$12,000+	$0	5-10 hrs onboarding	None
Claude API chatbot (10k msgs)	$15-40	$180-480	$500-1,200 (contractor)	8-15 hrs build	Full
Claude chatbot + n8n cloud	$45-65	$540-780	$500-1,200 (contractor)	8-15 hrs build	Full

The setup cost for the custom build is a one-time expense that is recovered within 2 to 3 months of Intercom savings and within 1 to 2 months of Drift savings. The “control” column matters more than it looks. When Intercom changes pricing (they do), you are stuck. When your custom bot needs a new FAQ added, you edit a text file.

What to Build First

Start with the narrowest version: a bot that answers your top 10 questions and collects a name and email before escalating to a human. That covers 70% of your current chat volume.

Week One: The Minimum Viable Bot

Ship the 10-question bot in week one. Set up the n8n webhook, write a focused 600-word system prompt, connect it to Claude 3 Haiku, and embed a basic widget on your site. Do not add features. Do not try to make it handle edge cases. Get it live and collecting real conversations.

Week Two to Four: Iteration From Real Data

Run the bot for 30 days. Then look at the conversation logs, which you own and which are stored wherever you configured them, and find the next 10 questions to add. This is where the custom build beats SaaS on a second dimension: you can read every single conversation and improve the bot based on real data. Intercom gives you analytics dashboards. Your custom build gives you the raw logs.

Escalation Routing Without a Fancy Inbox

For escalation, n8n can send a Slack message or email the moment a user asks something the bot cannot handle or explicitly requests a human. You do not need a fancy inbox to make this work. A Slack notification with the conversation transcript and the user’s contact information is enough for most SMB teams to respond within the hour.

One Thing People Get Wrong

They try to build too much context into the bot upfront. A system prompt that is 5,000 words of every policy, every product detail, and every edge case produces a confused, slow, and expensive response.

Keeping the System Prompt Tight

Start with 500 to 800 words of clean, direct context. The best-performing SMB chatbots are the ones with the most opinionated system prompts, not the most comprehensive ones. Tell Claude exactly what it is, what it should say, what it should never say, and when to hand off. Test it against your 10 most common questions before you go live. If it fails on more than 2 of them, the problem is almost always the system prompt, not the model.

Measuring Bot Performance

Track three numbers from day one: containment rate (percentage of conversations fully resolved by the bot without human escalation), escalation accuracy (percentage of escalations that genuinely needed a human), and cost per resolved conversation (total API cost divided by resolved conversations). These three metrics tell you whether the bot is working and where to improve it. No SaaS dashboard required.

The Bottom Line

If you are paying for Intercom or dealing with Drift pricing and your use case is mostly FAQ deflection and lead capture, you are 15 hours of build time away from cutting that bill by 80%. The Claude API is cheap, n8n handles the plumbing, and you end up with something you actually own and can improve.

The math is not close. At 2 seats, Intercom costs nearly 5 times more per year than a Claude-powered build. Drift costs 7 to 30 times more depending on the tier. For SMBs that have been accepting those bills as a cost of doing business, a custom build is not a technical project. It is a financial decision with a clear payback period and a better long-term outcome.

Start with the 10-question bot, get it live, and let the savings fund whatever you build next.