Last 40Networking Pass tickets — then sold out
Design Stage

Designing Fin – The #1 AI Agent for Customer Service

September 18, 2025 3:15 PM
Power Talk
🇬🇧 English
Wisła

About

What it really takes to design a leading AI product—and change the company in the process

Designing Fin wasn’t just about launching a new product. It meant tearing up 15 years of company habits, process, and structure, and starting again.

This talk tells the inside story of how we built Fin, Intercom’s AI Agent and the market leader in its category. I’ll dive into the design challenges that came with it: rapidly evolving technology, non-existent UX patterns, questions of trust and transparency, and the ever-present risk of hallucinations.

To meet that challenge, we didn’t just redesign the product—we redesigned how we build. We collapsed handoffs. We rewired how design, product, and engineering work together. And we replaced polished, linear processes with something faster, messier, and more experimental.

This is a story about design leadership, radical reinvention, and what happens when you stop designing interfaces—and start designing intelligence.

‍

Talk by

Photo of Thom Rimmer, Snr Director, Product Design at Intercom
Thom Rimmer
Snr Director, Product Design at Intercom

Watch the full talk

Watch this WaysConf session, then continue with related talks or explore the current programme.

Explore the 2026 agendaMore WaysConf recordings
From the recording

Talk in brief

In this 2025 case study, Intercom product design leader Thom Rimmer explains how the company responded to generative AI by building Fin and changing the organization around it. The product story is also an operating-model story: Intercom replaced stable product-area squads with focused workstreams, centralized design so people could move to the highest-priority work, and asked design leaders to stay close to product quality. Rimmer argues that AI shortens the path between an idea and working software, so designers need broader product judgment and more ability to build. He describes designers shipping reviewed code, prototyping inside the real product, and owning design improvements after release. The talk does not present this model as a finished formula. Its central lesson is that a company pursuing a disruptive product may need to let go of processes optimized for its previous business while retaining high standards for craft and execution.

Key takeaways

  1. 01

    A disruptive product may require a disruptive operating model

    Intercom chose to build an AI agent even though it challenged the seat-based model of its established support product. The organizational changes followed from that strategic conviction.

    Watch from 3:01
  2. 02

    Workstreams concentrated accountability

    A single directly responsible individual owned each priority, teams flexed around the work, and priorities were reviewed regularly instead of being tied permanently to product areas.

    Watch from 10:20
  3. 03

    Centralizing design made staffing more flexible

    Designers no longer belonged permanently to one squad. The central model allowed the team to move design capacity as new technical opportunities and product needs appeared.

    Watch from 12:17
  4. 04

    Design leadership stayed close to the work

    The talk distinguishes product leadership from process-only management: leaders set the quality bar, challenge safe choices, and help teams make stronger product decisions.

    Watch from 13:10
  5. 05

    Designers became builders, not only interface authors

    Intercom had designers set up development environments, ship reviewed code, and prototype with real product data—reducing the gap between design intent and delivery.

    Watch from 15:49
Read edited transcript highlights

These concise notes were edited from automatic captions and checked against the talk structure. They are not a verbatim transcript.

Fin as the catalyst

Rimmer introduces Fin as Intercom’s AI agent for customer service and explains the talk’s resolution metric: a conversation counts only when the agent resolves the customer’s issue rather than handing it to a person. He describes a product that can answer questions and, through integrations, complete tasks such as returns, refunds, and account changes. Those product capabilities frame the larger question of the session—why an established SaaS company decided it had to rebuild both its offer and the way it worked.

Watch from 0:35

Choosing change over protection

The old Intercom business connected customers with human support teams and was already commercially successful. The team nevertheless believed large language models matched many recurring support tasks and would reshape the category. Rather than protect the existing seat-based model, the company treated the new technology as an existential strategic shift. The talk emphasizes conviction under uncertainty: the team had a strong view of the direction, but not a fixed picture of the product six or twelve months ahead.

Watch from 3:01

Organizing around the most important work

Permanent product-area squads made sense for a mature application, but not for a new product whose needs changed with the technology. Intercom introduced workstreams with one accountable owner and staffing that could change with the problem. Priorities were reviewed, continued, paused, or replaced. Designers could move between workstreams instead of remaining attached to one product surface, concentrating design effort where the company believed it would have the greatest effect.

Watch from 10:20

Leadership, craft, and generalists

The design organization also changed what it expected from leaders and individual contributors. Leaders were asked to stay close enough to the work to define quality, challenge weak solutions, and encourage creativity. Designers still needed strong core craft, but the team placed more value on agency, adaptability, and the ability to learn adjacent skills. The goal was not rolelessness; it was a group capable of responding to changing product problems without waiting for a perfect handoff chain.

Watch from 12:17

Closing the gap between design and delivery

Rimmer describes a shift from producing flat interface representations toward building and testing working experiences. Every Intercom designer set up a development environment and shipped a production change through engineering review. The team then used AI-assisted tools and an internal prototyping environment to work with the company’s design system and realistic customer data. The intended benefit was a tighter feedback loop: ideas could be evaluated as functioning product behavior instead of only as polished screens.

Watch from 15:49

The playbook remains unfinished

The talk closes with a tentative model for future designers: help shape what is built, express direction through functional prototypes, execute with both speed and craft, and own continuous improvement after release. Rimmer is explicit that the best methods are not settled. His advice is to commit to meaningful change, let go of process that no longer serves the strategy, approach the opportunity with optimism, and value curiosity because the field is too new for anyone to rely only on past experience.

Watch from 21:29
Explore WaysConf 2026

View the agenda

Get tickets