How should we budget for seat-based helpdesk pricing plus usage-based AI resolution pricing (and avoid surprise costs)?

Most support leaders are now budgeting for a hybrid model—seat-based helpdesk pricing plus usage-based AI resolution pricing—and discovering that the complexity isn’t the line items, it’s the variability. You want the upside of AI resolution without getting caught by surprise costs when volume spikes, channels expand, or adoption grows faster than planned.

Quick Answer: Treat seat-based helpdesk pricing as your fixed “capacity floor” and AI resolution pricing as a variable, modeled from your conversation volume and target AI resolution rate. Put guardrails in place—caps, routing rules, and monitoring—so AI can scale with demand without blowing up your budget.

---

The Quick Overview

• What It Is: A budgeting approach for combining a traditional, seat-based helpdesk (agents, inbox, ticketing) with usage-based AI resolution (e.g. Fin AI Agent resolving customer queries across channels).
• Who It Is For: Support leaders, RevOps, and finance partners planning for an AI-powered helpdesk—especially those migrating from legacy tools or layering AI onto an existing system.
• Core Problem Solved: How to predict and control total customer service costs when human capacity is fixed by seats, but AI usage and conversation volume are variable.

---

How It Works

At a system level, you’re blending two cost structures:

• Fixed costs: Helpdesk seats (agents, admins), often on monthly or annual contracts.
• Variable costs: AI resolution events (e.g. successful Fin resolutions), typically billed per resolved conversation or per “AI unit” of work.

The goal is to design your operations so the variable piece behaves predictably. That means:

1. Baseline your volume and handle mix.
   Pull 6–12 months of data: total conversations, channels (web, email, WhatsApp, Instagram, SMS), time-of-day patterns, and the breakdown of simple vs. complex queries.

2. Model AI resolution scenarios.
   Estimate how many of those conversations AI (like Fin) can resolve at different resolution rates (e.g. 40%, 60%, 70%) and what that means for AI usage cost and human load.

3. Add operational guardrails.
   Use routing rules, Workflows, and channel policies to control which queries AI touches, and set monetary or volume-based caps so you don’t get surprised by a spike.

From there, you track real performance—resolution rates, handoff patterns, cost-per-resolution—adjust thresholds, and tighten the model each month.

---

How to Structure a Combined Budget (Step-by-Step)

1. Lock in your fixed “seat” baseline

Start with the part you know best: your helpdesk.

For Intercom, this is your Helpdesk + Inbox seat count:

• Frontline agents
• Team leads/supervisors
• Admins and workspace owners
• Occasional access roles (product, engineering, etc.)

Agree internally on:

• Required coverage: Time zones, languages, SLAs.
• Minimum viable staffing: The number of people you realistically need logged in if AI were offline.
• Growth assumptions: Expected growth in conversation volume and customer base.

This gives you your fixed monthly cost floor—your “no-regrets” spend that keeps service running even if AI is in limited rollout or being tuned.

2. Map your conversation volume into clear buckets

AI resolution pricing only becomes predictable if you’re clear about what AI will actually handle.

Segment your historic conversation data into:

1. High-volume, repetitive workflows
   Examples: password resets (when not fully self-serve), order status, shipping updates, billing dates, basic account questions, FAQs.

2. Policy- and procedure-driven workflows
   Examples: plan changes, refunds, cancellations, KYC steps, basic troubleshooting flows.

3. High-complexity or high-risk workflows
   Examples: outages, billing disputes, security concerns, escalations to legal, enterprise account changes.

In Intercom, I typically align these to topics (via tagging and AI Insights) and workflows:

• Topics where Fin AI Agent should be first line of resolution
• Topics where Fin should help but always hand off (e.g. collect details, summarize logs)
• Topics where Fin should not engage beyond maybe a standard “I’ll get you to a human” message

This mapping is the backbone of your cost model.

3. Estimate AI resolution volume and cost

Now convert those buckets into an AI usage forecast.

You need four inputs:

1. Monthly conversation volume
   Sum across channels you intend to expose to AI (Messenger on web and in-product, email, WhatsApp, Instagram, SMS, etc.).

2. Eligibility rate
   Percentage of conversations that you’ll route through AI. Example:

   • 70% of web and in-product Messenger conversations
   • 40% of inbound emails (where “Email To” is your support address, but “Email Cc” rules stop AI from replying in complex B2B threads)
   • 0% of certain channels if you want humans-only handling there

3. Expected AI resolution rate
   Start conservative. Intercom customers see an average Fin resolution rate of ~66%, which typically increases over time, but budget at a lower rate initially (e.g. 40–50%) to avoid surprises.

4. Per-resolution pricing
   This depends on your AI plan. Model it as “cost per AI-resolved conversation.”

Example model (round numbers):

• 50,000 total conversations/month
• 60% routed through Fin → 30,000 AI-eligible conversations
• 50% AI resolution rate → 15,000 resolved by Fin
• $X per AI resolution → 15,000 × $X = AI variable spend

Add your fixed seat cost and you have a blended cost view.

The key: run this with three scenarios—low, expected, and high AI resolution rates—so you understand the range of possible spend.

4. Design guardrails that control AI usage

To avoid surprise costs, you shouldn’t let AI touch “everything everywhere all at once.” Instead, configure guardrails at three levels:

a. Guardrails by channel

Decide where AI runs by default and where it’s constrained. For example:

• Web / in-product Messenger: Fin is the default first responder for eligible topics.
• Email: Use Workflows to limit Fin to cases where:
  • “Email To” matches your main support address
  • The conversation is first-party (not a vendor or CC-heavy thread)
• WhatsApp / SMS: Start with narrower coverage; expand once you trust performance.
• Instagram / social: Maybe restrict AI to FAQs and basic status updates initially.

This lets you throttle AI exposure in higher-risk channels while still benefiting from volume reduction where it’s safest.

b. Guardrails by topic and intent

Use Intercom Workflows + AI Insights:

• Route only certain intents to Fin (e.g. “billing cycles,” “shipping status,” “password help”).
• Set “no-Fin” rules for topics like refunds over a certain amount, security alerts, or enterprise account changes.

This protects your budget and your risk surface—Fin isn’t burning through usage on complex one-off scenarios that will escalate anyway.

c. Guardrails by caps and thresholds

Agree on:

• Monthly budget cap for AI usage
• Soft thresholds where you review usage (e.g., if AI spend hits 80% of budget by day 20, you tighten guardrails temporarily)
• Performance thresholds (e.g. if resolution rate drops below 40%, you pause expansion and tune content/procedures before exposing more volume)

Ask your Intercom account team which controls are available for caps or alerts on AI usage; operationally, you can also:

• Temporarily route more traffic directly to agents.
• Restrict topics Fin can address until you’re back within budget.

5. Bake in an improvement loop (so spending efficiency goes up)

One mistake I see: teams lock in a budget model and never revisit the assumptions. With a system like Intercom, AI performance isn’t static.

Build a monthly cadence:

1. Review AI Insights and reporting

   • Resolution rate by topic/channel
   • Top queries AI can’t resolve or is escalating
   • Cost per AI resolution vs. cost per human-handled resolution

2. Update your knowledge and procedures

   • Add or refine Help Center articles Fin can use.
   • Turn successful human responses into procedures that Fin Tasks can execute.
   • Clarify policies for edge-case questions.

3. Tune routing and guardrails

   • Expand AI coverage where resolution and CSAT are high.
   • Tighten or remove AI routing where escalation rate is high or cost-per-resolution is poor.

Over time, this self-improving loop means:

• AI resolution rate goes up.
• Human load per seat goes down.
• Your cost per resolved conversation drops—even if your raw AI usage increases.

---

Features & Benefits Breakdown

Core Feature	What It Does	Primary Benefit
Seat-based Helpdesk	Provides fixed, predictable costs for your human team with a configurable Helpdesk, Inbox, and ticketing.	Establishes a clear cost floor and guaranteed human capacity for complex issues and edge cases.
Usage-based AI Resolution (e.g. Fin AI Agent)	Bills only when AI successfully resolves a conversation across supported channels.	Converts a portion of your volume into variable, value-based spend that scales with demand.
Workflows & Guardrails	Control when and where AI engages (by channel, topic, customer segment, or email predicates like “Email To”).	Prevents runaway usage and focuses AI spend on high-fit, high-resolution scenarios.
AI Insights & Reporting	Surfaces resolution rates, gaps, and performance by topic/channel across AI and human support.	Lets you continuously optimize coverage and improve cost-per-resolution over time.
Fin Tasks/Procedures & Data Connectors	Allow AI to carry out multi-step actions (e.g. check order status, update fields) via API calls and business logic.	Increases the range of queries AI can truly resolve, boosting ROI on your AI usage.

---

Ideal Use Cases

• Best for teams modernizing from legacy helpdesks: Because you can lock in predictable seat-based Helpdesk costs, then add AI resolution gradually and safely—channel by channel, topic by topic—without taking on unbounded risk from day one.
• Best for high-growth digital products with spiky volume: Because usage-based AI resolution absorbs demand spikes (launches, outages, seasonal peaks) without you having to over-hire seats, while guardrails keep your AI spend inside a predictable band.

---

Limitations & Considerations

• Forecasting depends on historical data quality: If your current helpdesk doesn’t track topics, channels, or resolutions cleanly, your initial AI usage model will be rough.
  • Workaround: Start with conservative assumptions, then tighten your forecast after 1–2 months of real AI usage data.
• AI performance improves over time (not day one perfection): Fin’s average resolution rate is 66% across customers and tends to climb, but you shouldn’t bank your budget on best-case performance from the start.
  • Workaround: Plan a ramp period with lower assumed resolution rates and gradually shift volume to AI as you see actual results.

---

Pricing & Plans

Intercom’s pricing combines:

• Helpdesk seats: Fixed, per seat, usually priced per month (often discounted on annual plans).
• Fin AI Agent usage: Variable, based on AI resolutions (and potentially differentiated by plan or volume tier).

When budgeting, treat them as:

• Foundation: Seat-based Helpdesk cost for your minimum staffing and operations.
• Scaler: AI resolution spend that flexes with volume and the proportion of queries you route to Fin.

Two common planning patterns:

• Control Plan: Best for teams early in AI adoption needing tight budget predictability.

  • Lower AI coverage (fewer topics and channels), conservative caps, a smaller pilot group of seats.
  • Aim: prove value, learn where AI resolves best, and refine your model before broad rollout.

• Scale Plan: Best for teams with strong historical data and high confidence in AI.

  • Broader Fin coverage across Messenger, email, and select messaging channels; more seats leveraging Copilot and workflows.
  • Aim: maximize AI resolution rate and reduce time-per-conversation, while tracking cost-per-resolution and expanding guardrails intelligently.

Your Intercom account team can help map your volume, target resolution rate, and channel strategy into concrete pricing options and caps.

---

Frequently Asked Questions

How do I prevent AI resolution costs from spiking during a launch or incident?

Short Answer: Use routing rules and caps to control where AI is active, and treat high-risk windows (like launches or incidents) as special “playbooks” with tighter guardrails.

Details:
For major releases or incidents, it’s common to see 2–3× the usual volume. If you’ve exposed “everything” to AI, usage-based costs can jump with that spike. Instead:

• Define incident-mode Workflows where:
  • Fin handles only FAQs and status updates.
  • Complex queries route directly to agents.
• Consider temporarily narrowing channel coverage (e.g. restrict AI to Messenger, pause it on email during a critical window).
• Coordinate with your account team on soft caps or monitoring, so you know early if you’re trending above budget.

This approach lets you absorb volume spikes without taking on uncontrolled AI usage.

---

How do I know if I’m overspending on seats vs. under-investing in AI?

Short Answer: Track cost-per-resolved conversation across AI and humans, then shift budget toward whichever is reducing that blended cost while meeting your quality targets.

Details:
In Intercom, you can monitor:

• Resolution rate by topic and channel for Fin and agents.
• Average handle time and backlog trends for agents.
• AI vs. human share of total resolutions.

Signals you’re over-invested in seats:

• High agent idle time but low AI coverage.
• A large proportion of simple queries still being handled by humans.
• Rising volume but static or rising seat count.

Signals you’re under-invested in AI:

• Agents constantly at (or over) capacity.
• Long queues on simple queries that could be handled by AI.
• Topics where Fin already shows strong resolution rates, but only sees a fraction of eligible volume.

Shift budget by:

• Adding AI coverage (more topics/channels) where Fin already performs well.
• Holding or slowly reducing seat growth as AI takes on more volume, rather than hiring ahead of demand.

---

Summary

The cleanest way to budget for seat-based helpdesk pricing plus usage-based AI resolution pricing is to separate the two in your model—but operate them as one connected system. Seats give you a predictable capacity floor; AI resolution gives you scalable, variable capacity that flexes with volume.

If you:

• Baseline your volume and topic mix,
• Model AI resolution under conservative and optimistic scenarios,
• Put clear routing and guardrails around where AI engages,
• And review AI Insights monthly to adjust coverage,

you can avoid surprise costs while steadily improving your cost-per-resolution and customer experience.

---

Next Step

Get Started