How to add AI to B2B tools – 3 case studies

AI is coming to B2B – with or without you

It’s almost impossible to keep up with the pace of innovation that is happening in the AI space. Every week we’re seeing a flurry of new startups offering brand new capabilities. It started with Midjourney’s generative AI, then we got OpenAI's ChatGPT, and now text-to-video with Sora. And I’m sure that many more amazing projects will see the light of the day by the time you read this post.

All this (exciting) noise surrounding AI can give you a bit of fomo if you’re working on a “classic” type of platform. You know, one of these tools where users still need to click around to perform actions, write updates, move tasks to a completed state. The good old B2B software. It can feel like you’re being left out of an amazing transformation. Worse, it can feel like you’re about to be disrupted if you don’t figure out a way to join the movement.

I’m in the camp that thinks that business will need to treat AI like mobile. Some will benefit from being AI-first, but there's a huge number of other platforms that just need to understand how to be AI enhanced.

It would be wise to not delay that effort too much. Humans are incredibly good at digesting technology leaps and adjusting their expectations to the newly established standards.

  • 2 years ago, we were mocking generative AI for its inability to reproduce reality. Now we’re barely holding onto the belief that we’ll be able to see some weirdly shaped fingers (but will we?).
  • We complain about some mild hallucinations and mistakes in ChatGPT, and we seem to have completely accepted that it’s normal to converse with a machine and get 2-3k words analysis that make complete sense.
We quickly went from mocking the picture on the left to expecting the right to be a standard

It may seem like you have a lot of time ahead of you to figure things out, but I personally think that buyers' expectations will have shifted massively when 2025 starts. Users will favour tools that best increase their productivity, and your ability to leverage AI will become a standard evaluation criteria.

Today, we’re mostly fascinated by conversations around AI models and AI native apps like Devin, an autonomous AI software engineer. But, soon users will expect every B2C and B2B app to become AI enhanced, and the ones that do not manage to keep up will see their usage drop.

It might be about talking to your data, performing analysis, letting you visualise products in your own home (or on yourself), translating audio on the fly. The point is, you’ll have to figure out what makes sense for you.

To give you some inspiration I’m going to look at 3 B2B products that have added AI in very specific ways to give a much better experience to their users.

3 examples of AI applied in B2B services

Better product-led onboarding by generating contextual goals with B2B AI in Tability

A common challenge for self-serve products is activation. It’s not that hard to enable users to create accounts by themselves. But getting them to complete their setup, and making sure that they get a few Aha moments as part of their onboarding process can present various degrees of difficulty depending on the service that you offer.

A scheduling tool like Calendly won’t require much to get started, but in the case of Tability (a goal-tracking platform for teams), things can be a bit more complicated:

  • Many people signing up want first to understand the general goal-setting experience, and what kind of dashboard we have to offer
  • They may not have a clear set of goals in mind (they’re just exploring!)
  • But, you need goals to understand the value of Tability.

Our first approach was to build a library of OKR templates that people could pick during their onboarding. We did our best to cover as many functions as possible (sales, product, marketing, customer success…) but we quickly realised that this was (1) an expensive effort and (2) still not relevant to the user.

At best, we’d have your typical “OKR to achieve product-market fit” example that you could use.

But then ChatGPT came along, and after playing with OpenAI’s API we figured out a better way to onboard users: we could simply ask them about their general objective, and turn their response into a set of relevant goals using our own goal-setting AI.

B2B AI onboarding in Tability with goal-setting AI
Using AI to generate tailored goals during the onboarding

This kind of personalised onboarding is something that I believe many other tools will adopt. We’ve all been asked about our “role” when signing up for a tool, but how many times did this lead to having a tailored onboarding experience?

For Tability, the goal-setting AI is an integral part of the onboarding process, as well as something that sets the platform apart for users

AI now seen as a differentiator for Tability

Using B2B AI to automate release notes with released.so

Let’s be honest. There’s a significant number of teams that stop the definition of done at “it’s shipped”. And by “shipped” we mean that the code has been merged and is now in production.

But, how can the users know about the recent releases?

Some new features might be easy to spot, but many fixes and small improvements will go unnoticed unless you take the time to explain what changed. This is why we need release notes, but the process to gather the list of recent changes and turn them into a series of announcements can be time-consuming if not cumbersome:

  • You need to track back all the features/bugs/improvements worked on since the last release
  • You need to turn each ticket into a piece of couple of paragraphs explaining what was done
  • You need to have a way to publish that update and push it to your users

This feels like a lot of friction, and small teams will often be exhausted by the time they get to the notes. Knowing how to write code is not the same as knowing how to write a release note. Personally, I often suffer from writer’s block when I work on our announcement posts (I tend to overthink it).

So, cue in Released.so that started as a solution to automate the gathering-the-ticket step. It’s a Jira app that makes it super easy to compile the list of changes in a couple of clicks. So far this is a classic B2B app that will remove a lot of friction, but things get more exciting when you look at how they’re leveraging AI to significantly increase the value of their platform.

Released.so allows you to create templates that have “magic” AI fields. These fields will allow you to apply custom prompts to each of the issues and use the issue data to create adequate summaries of the work that has been done.

B2B AI in Released.so with AI-driven release notes templates
AI-driven release notes templates in released.so

You no longer need to write the notes yourself, you simply get the release notes AI to transform tickets into a changelog for you.

I find this fascinating because it adds an exciting step to Continuous Delivery pipelines as we could theoretically get released notes automatically published to our users as soon as the code hits production.

Saving time in post-production with Loom B2B AI-timestamped transcripts

We use Loom a lot at work. It can be to send short feature demos, to create tutorials, or simply to communicate async as a remote team. The core of Loom is what makes it amazing: click a button and you can record your screen instantly. Stop the recording and the video is readily available for sharing.

Loom’s core value proposition is that it has a very short time-to-share experience. You don’t need to wait for an app to open up or for a file to be compressed and uploaded. And as a result the arbitrage is often reduced to picking the easiest option between:

  • Typing words in Slack, or
  • Recording a 3-4min video

Ok, so where’s AI in that?

Well, one thing that is mildly annoying with video is that it’s not searchable. So when your recording is longer than 4-5 minutes, then it can be hard to remember later what you were talking about. And you’re also forcing people to go through the entire length of the video to find the interesting bits to listen to (I tend to ramble a bit sometimes).

B2B AI in Loom with timestamped transcripts
Loom's AI generated transcripts are mapped to the video timestamps

Loom solved that problem by doing 2 things:

  1. It automatically creates a transcript of your video
  2. It adds timestamps to match the video to the transcript

Once again the magic is in the UX. I get transcripts from my Google Meet recordings, but I can’t easily match them to the video. Loom, on the other hand, makes it super easy to scan the text and jump to an interesting part of the video (or scan the text and decide it’s not worth watching).

Don’t make AI a gimmick – find your real use cases

If there’s one take-away it is that you can’t just copy what others are doing. Whatever your product does, there must be unique ways that you can enhance the experience for your users. People won’t be satisfied if you just throw a gimmick into your app that doesn’t bring value.

A couple of questions to consider:

  • Is it making the experience more personalised?
  • Is it saving time?
  • Is this bringing new insights?
Author photo

Sten Pittet

Co-founder and CEO, Tability

Share this post
Weekly insights for outcome-driven teams
Subscribe to our newsletter to get actionable insights in your inbox.
Related articles
More articles →