---
title: "How support software gets chosen when buyers ask AI"
description: "By being named while support leaders re-shop for AI agents, with clear outcome pricing and proof of resolution rates that buyers and AI answers can check."
canonical: "https://underneath.agency/resources/customer-support-software-ai-search"
published: 2026-10-07
updated: 2026-10-08
publisher: "Underneath (https://underneath.agency/agent)"
entity: "https://underneath.agency/.well-known/entity.json"
---
Guide · AI search

# How do customer support software companies get chosen when buyers ask AI?

By being named, accurately, at the moment support leaders are re-shopping their helpdesk for AI agents, and then proving in a trial that the resolution rates hold. Almost every support team is re-evaluating its tools at once, and pricing is moving from seats to outcomes, so buyers have more questions than ever and fewer settled answers. That makes this category unusually open to vendors who are easy for AI assistants to describe correctly.

## The short version

1. [Gartner’s survey](https://www.gartner.com/en/newsroom/press-releases/2024-12-09-gartner-survey-reveals-85-percent-of-customer-service-leaders-will-explore-or-pilot-customer-facing-conversational-genai-in-2025) of 187 support leaders found 85% would explore or pilot customer-facing conversational AI in 2025, and [Gartner predicts](https://www.gartner.com/en/newsroom/press-releases/2025-03-05-gartner-predicts-agentic-ai-will-autonomously-resolve-80-percent-of-common-customer-service-issues-without-human-intervention-by-20290) AI agents will resolve 80% of common service issues without a person by 2029.
2. Among enterprise technology leaders interviewed by [Andreessen Horowitz](https://a16z.com/ai-enterprise-2025/), over 90% said they were testing third-party apps for customer support, which means they are shopping.
3. Pricing is changing under buyers’ feet: [Intercom](https://fin.ai/pricing) charges $0.99 per outcome for its AI agent, and [Zendesk](https://www.zendesk.com/pricing/) bills AI agents per automated resolution on top of per-agent seats. In [our pricing study](https://underneath.agency/research/ai-pricing-accuracy-study), Zendesk tiers that no longer exist showed up in AI answers.
4. The money is upmarket: at [Freshworks](https://s21.q4cdn.com/987526491/files/doc_financials/2026/q2/Freshworks_Q2_2026_Earnings_Call_Transcript.pdf), customers paying over $100,000 a year grew 25% and make up about 40% of total recurring revenue, while its helpdesk business grew 3%, a sign that share in this market is won by displacing incumbents.
5. [HubSpot](https://transcripts.platformaeronaut.com/transcripts/HUBS-2Q25-transcript), which sells a helpdesk as well as a CRM, told investors that clicks from AI assistants convert much better than organic search because buyers arrive deep in their research.

## Who signs for a helpdesk, and how much is a support account worth?

A head of support or customer experience leads the choice; IT, finance and security sign off on larger deals.

The buyer owns a cost center under pressure and a customer experience under scrutiny. [Zendesk’s CX Trends research](https://cxtrends.zendesk.com/) reports that 85% of customer experience leaders say customers will drop brands over unresolved issues, even on first contact, and that 74% of consumers now expect customer service to be available around the clock because of AI. Those two numbers explain the purchase: leaders need more resolutions without proportionally more agents.

The value of a customer depends on team size and, increasingly, on how much automated work it buys. Freshworks reported that customers contributing more than $50,000 a year now represent over 55% of its total recurring revenue. Its helpdesk business ended the second quarter of 2026 at $400 million in annual recurring revenue, growing 3% year over year, and the company expects low-single-digit growth for the year. In a market growing that slowly, new revenue mostly comes from winning accounts away from someone else.

Pricing is in transition. Zendesk describes its pricing as primarily seat-based, per agent per month, with AI agents billed by “automated resolutions,” meaning requests resolved without escalation to a person. Intercom prices its Fin AI agent at $0.99 per outcome, and $9.99 for each prospect it qualifies and routes to sales. For a buyer, that turns a familiar question (“how many seats do we need?”) into a harder one: how many conversations will the AI actually resolve, and what will that cost?

## Why is almost every support team shopping at once?

Because AI agents changed what a helpdesk is expected to do, and leaders are under pressure to adopt them.

Gartner’s survey found 44% of support leaders exploring a customer-facing voicebot, 11% piloting one and 5% already deployed. Gartner also expects the shift to cut operating costs by 30% by 2029. Those are forecasts, but they describe the brief that support leaders are receiving from their own executives.

Vendors are answering with resolution claims, all self-reported and measured differently:

- Intercom says Fin averages a 76% resolution rate across 12,000+ customers and handles 2 million resolutions a week.
- HubSpot said its customer agent had 4,000 customers in mid-2025, with resolution rates averaging 55%.
- Freshworks said its AI agent’s deflection rates average 50%, reaching as high as 80% for mature deployments.

Each vendor counts a “resolution” its own way, so buyers ask someone to compare them. Increasingly that someone is an AI assistant, followed by a trial.

## At what point do support leaders bring AI assistants into a helpdesk search?

At the comparison stage, and the published signals come from vendors who sell support software.

HubSpot’s chief executive told investors in August 2025 that organic search traffic is declining as AI Overviews give answers, that the company is now cited by AI assistants more than any other CRM, and that clicks from AI assistants convert much better than organic search because buyers are already deep in their research. The same call noted that over 20,000 HubSpot customers had used its ChatGPT and Claude connectors. HubSpot sells to the same mid-market teams that buy helpdesks, so its experience is the closest public signal for this category, though it reports it for its business as a whole.

Outside helpdesks, [G2’s 2026 survey](https://company.g2.com/news/g2-research-the-answer-economy) of software buyers in general found 51% of buyers now start research with an AI chatbot more often than with Google. G2 sells to software vendors, so read it as a direction rather than a measurement for helpdesks specifically. Tools sold to the same teams are covered in [how collaboration software wins AI-first teams](https://underneath.agency/resources/collaboration-software-ai-search).

## Which questions do support leaders ask AI assistants?

Questions about switching, AI resolution, total cost, integrations and fit for their team size.

We wrote the prompts below to show how a head of support might phrase a helpdesk search; none was taken from a real buyer:

- Switching: “Best Zendesk alternatives for a 40-agent B2B support team that wants AI agents included.”
- AI performance: “Which helpdesk AI agents actually resolve billing questions without a human?”
- Cost: “How much would Intercom Fin cost us at 30,000 conversations a month compared with Zendesk?”
- Integrations: “Which support platforms integrate natively with Salesforce, Jira and Shopify?”
- Fit: “Best helpdesk for an ecommerce brand with 5 agents and seasonal spikes.”
- Risk: “Which customer service platforms are HIPAA compliant and let us keep data in the EU?”

Cost questions are where answers are most likely to go wrong, because outcome pricing is new and older pages still carry older prices. In our pricing study, 39 of the 64 prices that differed from the vendor’s page were found on another page of the vendor’s own site, including Zendesk’s article on its 2023 pricing update.

## How does a helpdesk named in an AI answer become a signed contract?

The assistant puts you on the shortlist; a trial proves the resolution rate; the contract follows.

**Shortlist.** A support leader asks for alternatives or a comparison and gets a short list with reasons. If your product is not on it, the evaluation proceeds without you.

**Trial.** This category lets buyers test before talking to sales. Intercom offers a 14-day trial with no credit card and unlimited Fin outcomes during it; Zendesk also runs a 14-day trial. A buyer can point the AI agent at its own help center and see how many questions it resolves, so a good trial converts the claim into evidence.

**Demo and procurement.** Larger teams add security review, integration checks and a migration plan. Freshworks’ figures show where the value sits: customers paying over $100,000 a year grew 25% year over year. Those deals rarely show up as a single AI referral; we infer they appear as direct visits, branded searches and demo requests from a buyer who already knows your name. Sales software vendors face the same pattern, covered in [turning AI answers into sales pipeline](https://underneath.agency/resources/sales-software-pipeline-from-ai-search).

**Expansion.** Outcome pricing means revenue grows as the AI resolves more conversations, so the first deal is the start of usage-based growth rather than a fixed seat count.

## Why do assistants name some helpdesks and not others?

The assistants explain little; the research favors accurate pages, independent comparisons and current, structured information.

**What Google has said.** Google’s description of AI Mode includes a [“query fan-out” technique](https://blog.google/products/search/ai-mode-search/) that issues multiple related searches across subtopics, then combines the results. A question about Zendesk alternatives can pull in pricing pages, review sites, comparison articles and help-center pages at once.

**Observed in our studies.** In [our study of self-ranking “best of” lists](https://underneath.agency/research/self-promoting-best-lists-study), 24.2% of AI-cited numbered lists with an identifiable publisher ranked that publisher first, and Zendesk was among the most frequent publishers of such lists, with 3. We found no statistically detectable difference in how often answers named the top pick of a self-ranking list versus an independent one, so these lists neither clearly help nor clearly hurt. In [our study of agent-readable websites](https://underneath.agency/research/agent-readable-web-study), only 3.2% of top websites returned a plain text version when an AI agent asked for one; zendesk.com and intercom.com were among the early adopters.

**Observed in other research.** A study of [B2B software citations](https://arxiv.org/abs/2509.10762) using 70 industry-targeted prompts found the page qualities most strongly associated with being cited were metadata and freshness, clear page structure and structured data. That is a correlation, not proof that adding them causes citations.

**Trust factors specific to support software.** Buyers weigh how a vendor measures resolution, which integrations work out of the box, how migration is handled, and whether security and compliance claims can be verified. A reasonable expectation is that assistants answering those questions lean on whatever is public: your pricing page, trust center, integration directory, help center and independent reviews.

## What does a helpdesk vendor lose when an assistant leaves it off the list?

The switch you never heard about, in a market where switching is how share moves.

When a helpdesk market grows in the low single digits, as Freshworks expects for its own helpdesk business, the main source of new revenue is accounts leaving an incumbent. Those moments are now frequent, because AI agents are prompting re-evaluations across the market. We infer that an AI answer that omits you, or quotes your old pricing, costs you the shortlist at exactly the time buyers are most willing to move. Vendors with wrong prices in AI answers face a second cost: a buyer who sees an outdated figure may rule you out before visiting your site. We look at how falling search clicks affect pipeline more generally in [our article on AI answers and pipeline](https://underneath.agency/resources/ai-answers-pipeline-revenue).

## What should GEO cover for a helpdesk or AI agent vendor?

Your pricing, resolution claims and integrations, made easy for assistants to describe correctly, with no guaranteed recommendation.

For a helpdesk vendor, generative engine optimization (GEO) tends to come down to six pieces of work:

1. **Pricing that explains itself.** Publish one current page that states seat prices, outcome prices and what counts as a resolution, and retire or redirect old pricing articles. When an assistant still quotes a retired tier, follow the steps in [how to fix wrong brand information in AI answers](https://underneath.agency/resources/fix-wrong-brand-information-in-ai-answers).
2. **Resolution claims with a method.** State how you measure resolution or deflection, on what kind of traffic, and link to customer examples. Unexplained percentages are hard for buyers, and assistants, to compare.
3. **Honest comparison and migration pages.** Buyers ask switching questions, so publish fair “from Zendesk” or “from Freshdesk” migration guides and comparisons. Whether such pages earn citations is weighed in our [comparison pages](https://underneath.agency/resources/do-comparison-pages-help-b2b-ai-citations) article.
4. **Independent coverage and reviews.** Earn places in the comparison sites, analyst notes, newsletters and communities support leaders read, and keep review profiles current. Our article on [which “best of” lists matter](https://underneath.agency/resources/best-of-lists-ai-recommendations) helps pick targets.
5. **A help center built to be read.** Support vendors know this discipline better than anyone: Gartner found 61% of support leaders have a backlog of knowledge articles to edit. Apply the same standard to your own documentation, integration pages and trust center, with clear structure and visible update dates.
6. **Measurement by buying stage.** Track switching, AI-performance, cost, integration and compliance questions across ChatGPT, Gemini, Perplexity, Copilot and Google, and connect the results to trials, demos and pipeline.

## What remains unmeasured about AI search in helpdesk buying?

Nobody has published how AI visibility shifts win rates or contract value for helpdesk vendors.

Most of the numbers here are vendor-reported, including every resolution rate, and they use different definitions. The Gartner figures are a survey and a forecast. HubSpot’s comments describe its business as a whole, not only its helpdesk. G2 and Andreessen Horowitz describe software buying broadly, not this category. Our own studies record what AI answers cite and say about vendors such as Zendesk and Intercom, not which helpdesk a buyer then signs with. Treat AI assistants as an established part of the support software shortlist, with the size of their effect on revenue still unmeasured.

## How can a helpdesk vendor see which AI answers are costing it trials?

Compare what assistants say about your pricing, resolution rates and alternatives with what your trial data shows.

Run the switching, cost, AI-performance and integration questions your buyers ask through the main assistants and Google’s AI features, then match the answers against your trial and demo records to see which wrong or missing answers cost you evaluations, seats or outcome-priced volume. To have us do that work alongside your team, [talk to us about a helpdesk visibility audit](https://underneath.agency/contact). The follow-on work, including pricing pages that explain outcome charges and resolution claims stated with their method, is described on our [generative engine optimization service](https://underneath.agency/services/generative-engine-optimization) page.

## Frequently asked questions

### Do AI assistants get helpdesk pricing right?

Often not fully. In our study of software prices quoted by four assistants, many of the differences came from older pages on the vendor’s own site, and outdated Zendesk tiers were among them. Outcome-based pricing adds a new way to be misquoted, so it pays to check.

### Should we publish our AI agent’s resolution rate?

Yes, with the method. Buyers compare resolution claims across vendors, and a number without a definition is hard to trust or compare. Say what counts as resolved, on which kinds of conversations, and show examples.

### Are “best helpdesk software” lists on our own site worth it?

They are common in this category and AI engines cite them. Our study found about one in four cited numbered lists ranked their own publisher first, with no detectable difference in how often answers named that pick. A fair comparison page is a safer investment.

### How do we know if AI assistants are sending us trials?

Watch AI referrals into your trial and demo pages, ask new signups how they heard about you, and recheck how often assistants name your helpdesk for buyers’ questions. Then trace those accounts to the seats and resolution volume they end up buying.

## Sources

- Gartner (2024), [Gartner Survey Reveals 85% of Customer Service Leaders Will Explore or Pilot Customer-Facing Conversational GenAI in 2025](https://www.gartner.com/en/newsroom/press-releases/2024-12-09-gartner-survey-reveals-85-percent-of-customer-service-leaders-will-explore-or-pilot-customer-facing-conversational-genai-in-2025)
- Gartner (2025), [Gartner Predicts Agentic AI Will Autonomously Resolve 80% of Common Customer Service Issues Without Human Intervention by 2029](https://www.gartner.com/en/newsroom/press-releases/2025-03-05-gartner-predicts-agentic-ai-will-autonomously-resolve-80-percent-of-common-customer-service-issues-without-human-intervention-by-20290)
- Andreessen Horowitz (2025), [How 100 Enterprise CIOs Are Building and Buying Gen AI in 2025](https://a16z.com/ai-enterprise-2025/)
- Freshworks (2026), [Q2 2026 earnings call transcript](https://s21.q4cdn.com/987526491/files/doc_financials/2026/q2/Freshworks_Q2_2026_Earnings_Call_Transcript.pdf)
- HubSpot (2025), [Q2 2025 earnings call transcript](https://transcripts.platformaeronaut.com/transcripts/HUBS-2Q25-transcript)
- Intercom (2026), [Fin pricing](https://fin.ai/pricing) and [Fin](https://fin.ai/)
- Intercom (2026), [Intercom pricing](https://www.intercom.com/pricing)
- Zendesk (2026), [Zendesk pricing](https://www.zendesk.com/pricing/)
- Zendesk (2026), [CX Trends 2026](https://cxtrends.zendesk.com/)
- G2 (2026), [In the Answer Economy, Don’t Win the Click — Win the Answer](https://company.g2.com/news/g2-research-the-answer-economy)
- Google (2025), [Expanding AI Overviews and introducing AI Mode](https://blog.google/products/search/ai-mode-search/)
- AI Answer Engine Citation Behavior: Bringing the GEO-16 Framework in B2B SaaS (2025), [AI Answer Engine Citation Behavior An Empirical Analysis of the GEO16 Framework](https://arxiv.org/abs/2509.10762)
- Underneath (2026), [software pricing accuracy](https://underneath.agency/research/ai-pricing-accuracy-study), [self-ranking “best of” lists](https://underneath.agency/research/self-promoting-best-lists-study) and [agent-readable websites](https://underneath.agency/research/agent-readable-web-study)

---

This is the Markdown twin of https://underneath.agency/resources/customer-support-software-ai-search. The HTML page is canonical. Publisher: Underneath, https://underneath.agency/agent. Site index: https://underneath.agency/llms.txt.