All Articles

AI Agents v2

Co-Founder & CEO

Releases

Jun 10, 2025

Today we're releasing our biggest product yet...

...the first AI Agent built for B2B support.


Since starting Pylon in 2022, we've observed that existing AI support products do not work well for B2B.

Why? They focus too much on end-to-end ticket deflection. This is great if you're a B2C company with high volume, transactional tickets.

But if you're a B2B company, your support is...

→ more complex
→ involves multiple teams
→ requires context from more sources
→ is part of an ongoing relationship
→ has a long-tail of questions

B2B teams need a different type of AI agent, so we've spent the past 6 months building one.

Here's what's different...

𝟭/ 𝗙𝗼𝗰𝘂𝘀 𝗼𝗻 "𝘁𝗶𝗰𝗸𝗲𝘁 𝗽𝗿𝗲-𝘄𝗼𝗿𝗸" 𝘃𝘀 𝗳𝘂𝗹𝗹 𝗿𝗲𝘀𝗼𝗹𝘂𝘁𝗶𝗼𝗻

Today B2B teams rightfully don't trust AI to fully resolve the majority of their customer issues. Instead of focusing only on resolution, our agents can handle the parts of tickets that involve common, low-risk work like internal routing, asking clarifying questions, fetching internal customer context, and taking human-assisted actions. If a ticket cannot be fully resolved, at minimum all the pre-work should be done before your team sees the ticket.

𝟮/ 𝗙𝗹𝗲𝘅𝗶𝗯𝗹𝗲 𝗲𝗻𝗼𝘂𝗴𝗵 𝘁𝗼 𝗵𝗮𝗻𝗱𝗹𝗲 𝟭𝟬𝟬𝘀 𝗼𝗳 𝗾𝘂𝗲𝘀𝘁𝗶𝗼𝗻 𝘁𝘆𝗽𝗲𝘀

Given the long-tail of questions B2B support includes, our agents allow you to mass produce scenarios and workflows that we call "runbooks". You'll create one runbook per scenario (e.g. "I'm having trouble connecting my CRM") which will be capable of engaging with the customer, calling APIs, and performing all sorts of pre-defined actions.

𝟯 / 𝗦𝗲𝗹𝗲𝗰𝘁𝗶𝘃𝗲 𝗮𝗯𝗼𝘂𝘁 𝘄𝗵𝗲𝗻 𝗔𝗜 𝗴𝗲𝘁𝘀 𝗶𝗻𝘃𝗼𝗹𝘃𝗲𝗱

You'll want to gate which types of issues and customer segments your AI gets involved in. Most commonly we've seen our customers use account tiers to segment lower-paying customers to have more AI involvement rather than less.

𝟰 / 𝗔𝗽𝗽𝗿𝗼𝗮𝗰𝗵𝗮𝗯𝗹𝗲 𝗳𝗼𝗿 𝗻𝗼𝗻-𝘁𝗲𝗰𝗵𝗻𝗶𝗰𝗮𝗹 𝘁𝗲𝗮𝗺𝘀

We want all of our customers to be able to build AI agents, without additional technical resources. So runbooks are written in plain english with no code required (unless you want to).

𝟱 / 𝗢𝗻𝗲 𝗮𝗴𝗲𝗻𝘁 𝗶𝘀 𝗻𝗼𝘁 𝗲𝗻𝗼𝘂𝗴𝗵

Different teams can build, own, and operate their own agents and runbooks. Agents can work together or be used completely independently.


Finally... as you'd expect we're including the entire AI suite as part of this release including:

→ AI testing / simulation sandbox
→ AI performance measurement
→ AI attempted issue logs

--

This is one of the biggest products we've built at Pylon, and by far one of our more ambitious and differentiating from existing incumbents.

The team has crushed this project. Shoutouts to Aashish, Anish, Bennett, Bryan, Charleen, David, Dre, Wendy, Yoona.

Many fast follows are coming.

Get started today

We'll walk you through how you can get started and provide recommendations on how to scale your team and setup.