Trends, opportunities, pitfalls: AI revenue tooling

OpenAI released ChatGPT just seven months ago and at this point, you’ve probably seen maybe one too many “STOP MISSING OUT: Fifty ways successful people are using ChatGPT to save time → a thread” posts. That being said, it couldn’t be a more exciting time to be working in tech with a front row seat to the ripple effects of a development that’s going to fundamentally change the ways that we work and live.

Since November, we’ve chatted with 100+ pretty incredible operators (account executives, marketing leads, CX operations, VPs of revenue operations, sales enablement folks, etc) and early-stage founders building cutting-edge tooling and it’s been pretty fascinating watching everything unfold.

Growing pains that have always accompanied the development of most new categories and disruptive market forces persist as new tooling scrambles to pinpoint and satisfy customer needs and preferences — this seems especially amplified in one of the most challenging macro climates in quite a while.

Trends at the intersection of AI-powered tooling and revenue

Incumbents have the upper hand (sort of)

Everyone has eyes on the Gong vs Clari vs Outreach vs Salesloft battle right now, and having a massive data set is certainly an advantage in rolling out new features fast. But now that the low hanging fruit is kind of gone, it’s probably worth thinking about where it doesn’t make sense for these bigger players to compete - usually where it’s too risky or out of scope.

Too many buzzwords, too little time

Sometimes, you’ll end up on a site with a .ai domain and struggle to figure out how exactly the product would be different if it had a .com domain. There’s not too many synonyms for “workflows supercharged by AI” and so tens of companies with very different products will use it to describe very different things.

AI companies will need to lean into outcomes-based messaging and perhaps benefit from investing in more educational content, in-depth case studies and product reveals than a traditional software company might have to at a similar stage.

Privacy and functionality concerns

There are legitimate questions around security and how well certain tools work, but this is also beginning to get offset by the rise of interactive sandboxes that let prospects get to a magic moment without having to talk to InfoSec and infrastructure that enables internal engineering teams to deploy their own custom, secure AI models relatively easily.

Unique challenges newer AI revenue tooling and features face:

The technical investment and build-out for a truly dependable, usable, 10x tool that is genuinely AI-native is extremely challenging, especially for mid-market and enterprise.

Fitting into a user’s flow has always been incredibly difficult but critical — most folks simply aren’t going to explore random features of the software their manager’s manager bought seven months ago in the minimal down time they have between discovery calls and crunching data in Excel cells. Learning curves for AI features or products has to be optimized, and the value delivered has to be worth that heftier upfront investment.

Handling the sheer volume of data, much of which is unstructured and/or not clean, that the average Series B company has is a feat because of how large tech stacks have gotten. A number of deceptively simple AI products are actually incredibly complex to build and maintain.

For example, an English to SQL tool might seem like something you could duplicate by typing into chatGPT, but it doesn’t really work that way when you’re supporting spreadsheets with hundreds of rows and powering thousands of queries — S/O to our friends at Askedith.ai who are crushing it.

Going full stack very early because you have to, but before you can deliver reliably.

This isn’t exclusive to AI, but we think it’s amplified because of the technical hurdles.

The days of relaxed corporate credit card usage and folks dropping 20 grand on single-use point solutions are very much over. All-in-one suites need to nail messaging and the right set of features or find a pain point so underserved it can survive the bloodbath that is procurement as a point solution.

Greenfield opportunities: what we’re personally really excited about

Enabling teams who own a metric’s success to 10x their ability to make an impact without special knowledge or any technical expertise

A good example of this is scoring - leads, upsells, churn, etc. Traditionally, this might look like the software takes a “first stab” at what inputs you should care about to calculate the score. You’re then in the drivers seat adjusting inputs based on experimenting and gut.

As Sara Du alluded, for some bold teams the next iteration of this is autopilot, or at the very least the system itself being smart enough to recommend suggested changes to weights and inputs based on holistic returns you can then veto or approve.

Another potential evolution is that instead of waiting until you’re a Series D company (and even then, the data team will only handle hardcore models for a few critical parts of the business - and this is still somewhat siloed from GTM and business teams that have full business context), you can get access to the power of a ML team whenever your data is ready for it.

• Think account scoring, but taking into account all of your existing intent vendors and historical data of which prospects have reached a call with a solutions consultant vs those that dropped out after a few follow-ups, categorized by different segments and user personas.

Cutting down brutal implementation timelines and investment

Anyone who has experienced Salesforce knows that there’s often a decent amount of money and time invested into on-ramping, gaining knowledge, maintenance, etc. Same goes for launching most new, cross-functional data initiatives that involve stitching together multiple tools.

• In some use cases, you can actually abstract away some of the most tedious parts of onboarding and change management by handling setup for the user automatically or turning configuring otherwise complex UX into a series of English commands.

Rethinking anything that is considered a given right now

The cool thing is that a lot of what feels painful now (that previous tools couldn’t fully solve) are now back in the game. Everyone is shipping new product and features faster. Great article that dives deeper here.

• Whether you’re dreading buying another Tableau license, trying to hire unicorns who are both strategy and systems oriented, or are looking to cut your tech stack in half (you’re not alone in that one) — good days are ahead. Such is the nature of technology :)

If you made it to the end of this article, thanks for reading and definitely reach out if you’d like to chat about the future of GTM and revenue operations! Special shoutout to all the folks who we’ve been lucky to connect with in the last few months who have taken time out of their busy schedules to share their experiences.