Systems of Record vs. Systems of Action: Navigating the Future with AI-Powered Applications

Credit goes to my MIT classmate, Manu Lohiya, for introducing me to this framework. After reviewing our demo environment, he gave us some solid advice: “Systems of Record don’t cut it anymore, everything has to be a System of Action today.”

Well I’m glad he explained it to me, because there isn’t a ton of high quality content on this framework on the internet. We’re publishing this for posterity so we can pay it forward.

How to tell them apart?

If you’ve worked in enterprise tech, you’ll likely grok this case example very quickly.

Salesforce - the classic System of Record

The archetypical CRM and System of Record. It’s a log of all of your interactions with a customer, customer account properties, and helps you manage your sales pipeline. It’s an absolute must-have piece of software for serious sales organizations. But it has its limitations. Data quality is an issue, relying on the sales rep to enter data manually. They’re often busy, have incentives to obscure or omit granular details, and data gathered is ultimately biased by the perspective of the human doing the work before ending up in Salesforce.

Salesforce is a reporting database with self-reported data. As a sales leader, product marketer, or product manager, if I wanted to gain deeper insights into what’s happening with our customers, I would have to interview sales reps, sit in on sales calls, or setup dedicated environments for engaging customers like Customer Advisory Boards.

Gong - an early pioneer System of Action

Gong takes a different approach, incorporating ML from the earliest days of the company. Instead of relying on self-reported data, the app joins every sales call and records the video and provides a transcription of the meeting. This helps the Sales Rep stay focused instead of taking notes, but it also gives the opportunity apply ML to a large dataset, the dataset of every customer interaction. The software proactively looks for insights across all sales conversations: helping identify winning sales techniques from top performers and providing coaching opportunities for others, extracting insights from topics such as pricing negotiations, and alerts stakeholders when issues arise.

By granting Gong access to video calls (and now phone, email, etc.), Gong customers are now able to have an AI/ML solution that proactively utilizes data to drive a better outcome at the individual sales rep level as well as at the overall revenue level for an organization. And the key difference here is that the System of Action does most of the heavy lifting for this work automatically.

What kind of business are you building?

We’re in a transitional period where the incumbents are most likely Systems of Record. Even tech darlings like Notion are still largely Systems of Record. Talk to enough Notion users and you’ll see a primary selling point is the ability for the USER to infinitely customize and design their Notion spaces. Sounds a lot like a fun-to-use relative of Salesforce in that context, a great place to store and curate information manually.

This video was shot in late 2022, how do you think Francesco’s predictions have held up?

A key question Manu posed to us that afternoon was, ‘are we building a System of Record or a System of Action?’ If we’re building a System of Action, from Day 1, we have to think about the implications for our architecture, data model, and our process for product management decisions. Incumbents will be at a disadvantage to fully embrace what LLMs can do, and that is a startup’s opportunity.

If you believe LLMs have fundamentally changed the game, then I would argue your business must fully embrace it as a disruptive change in how you think about building a product.

How this changed us at Multiplayer

As Multiplayer evolves, we constantly ask ourselves why traditional task management and execution has failed. We listened for the challenges with task saturation, the effort and discipline it takes to maintain a personal productivity system, and what death by a thousand notifications feels like. There are more than enough to-do lists, project management tools, and cute UIs. A piece of paper works fine if we’re just storing information… a simple list of what to do next.

That’s why we use the lighthouse of a human assistant. What would you expect or hope a human assistant could do for you? We ask assistants to keep us on track, filter out the noise, prioritize what we focus on, and handle delegated tasks. These are actions. This is a different type of product altogether. This is what it takes to build something new today.

This post was not written with AI.