How to build your team to capitalize on AI

As Ramp strives to save customers more time and money, we have also set the ambitious internal goal to become the most productive company in the world. This mission requires us to persistently hone the role each Ramp team member plays down to its essence, cutting out low value work, and better listening to and delivering for our customers. AI plays a central role in helping us get there, but how?

I lead operations at Ramp and recently spoke at the Make with Notion conference about how we deploy AI across every member of the team. I shared steps companies can take to start using AI more effectively, especially in tasks beyond coding and software, and how tools like Notion have helped support that mission at Ramp.

I firmly believe the limiting factor in capturing the AI opportunity in most organizations is no longer the power of AI models but the empowerment of individual employees to capitalize on what they offer. At Ramp, we enable each of our employees to build AI tools tailored to their personal needs. We want every team member to be a builder, not just a “button pusher” who is stuck using tools created by others, so we can reimagine how work gets done and move faster.

Here are four tips I shared on how to create a “company of builders”:

1. Give AI prompting tools to everybody in your company, and enable them. At Ramp, 100% of our team has access to ChatGPT Enterprise, Notion (with Notion AI agents) and Perplexity, three tools that require their users to write actual AI prompts. We invest heavily in training every employee to use the tools during onboarding, and to work directly with departments around the company to activate these tools for their team’s needs. We meticulously tracked tool usage as we got started, since the only way for our teams to become great users of AI is by using AI. We now have 90% monthly active usage of Notion AI and nearly 100% weekly active usage of ChatGPT across our team - that’s hundreds of thousands of prompting practice sessions every month. Don’t deprive your team of this learning curve.

2. Overcome the vagueness epidemic and embrace precision: Tools like Google Search set the standard for the SaaS era. They encourage starting from a vague understanding of intent (a few words in a search bar) and sifting through the results to find what you need. AI tools, in stark contrast, demand precision—the more context you give in a prompt, the higher the quality of the output will be. But realize you can use AI to help you be more precise by prompting it to ask you questions, then feeding those answers directly into more thorough prompts. For an extra shortcut, AI tools like meeting note takers (Notion AI Meeting Notes, Granola) and voice-to-text dictation tools (Wispr Flow, superwhisper) can ensure that more context is available from the start to drive greater precision with less work.

3. Take knowledge management much more seriously: . Your AI tool performance relies on the accuracy of your knowledge, and that accuracy will soon be tested millions of times per day. A+ knowledge in the AI era means four things:

• Consolidating to a single source of truth
• Building feedback loops that maintain accuracy
• Describing knowledge with crystal clarity
• Identifying what knowledge your team needs for a given task.

Notion has been a major boon in several of these goals. By allowing Ramp to consolidate half a dozen former information storage tools into one (wikis, product roadmap, task management, operating procedures, customer research, sales and marketing docs), Notion has become the single source of truth, a position further solidified by Notion AI’s integrations to other core Ramp information sources like Slack, Github, and Linear. Notion’s AI agents then power multiple feedback loops that identify missing, incorrect, or confusing information and facilitate accurate updates.

The ability for Notion AI users to select the sources the platform searches for a given task also encourages more knowledge precision from the team. This is something Notion’s increasingly personalized agents can further take advantage of to ensure, for instance, that we aren’t answering product questions with information that’s years out of date.

4. Deploy tools that let non-engineers build full AI workflows: With great prompting and knowledge skills, team members can use AI to complete individual tasks with high quality. Workflow tools help to deploy those task-level innovations automatically at high scale. Historically this has been the domain of engineering, but new AI tools democratize this work. At Ramp, we use tools like Gumloop and Notion Custom Agents to enable each team to push AI-powered innovations within their own domain. Marketers are automatically running copy through AI editors, product teams automatically update project context, and operators are building nuanced flows where AI attempts to automatically answer questions or add context at every step of the way.

Customers win too. We’ve built tools to instantaneously compare customer feedback with our product roadmap, and close the loop if we see that a feature is launching that will address a customer’s issue. And Ramp Assist, our in-app support agent, successfully resolves over 90% of requests without needing to wait for the next human support specialist to be available.

AI operations is still a young discipline, and there is a lot for everyone to learn. But the possibilities of what can be built are expanding every couple of months, and the time to start using AI is now. The moment demands a dose of humility — there’s no reason to believe the specific tasks we’ve mastered in the last generation of work are particularly necessary in the next. But the moment also rewards curiosity – it’s fun to wake up, pick something you don’t like about your work, and build your way out of it. This innovation compounds — one seller or operator or designer who has a breakthrough can bring infectious enthusiasm - and productivity - to the rest of their team.

Begin putting everybody in a builder’s seat and ask them: what will you build first?

Photos courtesy of Boris Zharkov (@boriszharkovstudio).