Buzzword bingo

If you are an entrepreneur or business leader, you have likely found yourself in brainstorming sessions where the goal was to find ways to use GenAI in your business.

You were overwhelmed with ideas that sound exciting and could reinvent every aspect of the company. Sprinkle a few buzzwords like agentic, operator, multimodal, and AGI, and you’ll end up no closer to a decision than you were at the beginning of the meeting.

Granted, every company needs a few AI moonshot projects that reinvent the business. But before going for the Olympics, it needs to develop “AI fitness.” In today’s environment, this means launching practical wins within 3-month production cycles.

AI as the intern

Instead of asking what AI can do, I’ve found it helpful to ask the team: if you had a large number of free, reasonably educated interns, what would you ask them to do?

When we visualize AI agents as interns, our mindset shifts from concepts to tasks. We imagine how we’d give detailed instructions to the interns and what specific tools we’d give them access to.

We have more realistic intuitions about scenarios where the interns can save us time (or save our customers’ time), but we can also leverage our experience of the many stories when interns have taken us more time than if we just did things ourselves.

AI agents, like interns, are particularly effective at:

• Doing extensive research and summarizing findings in a predefined format.

• Cleaning data or transferring it from one platform to another.

• Executive repetitive processes, once their supervisor has performed the work enough times to know how to break them down into individual tasks.

• Monitoring the industry or the business and escalating promising or unusual developments to the supervisor.

By contrast, AI agents are not (yet) very good at:

• Getting things done when goals or processes are unclear.

• Coordinating people (either customers or internal teams).

• Interacting in real-time with high-value customers or stakeholders.

• Hustling and taking extreme ownership for business outcomes.

Also, you don’t want an intern or an AI agent to keep a lot of information in their own ad-hoc notes and memory. You’d much rather have them populate the systems and databases already used by the business, so that they augment your existing processes instead of creating new silos.

Evaluation framework for GenAI ideas

When you ask your team to generate GenAI business ideas, request that they answer the following questions for each idea:

• What is the job title of the intern?

  • E.g., social media intern, lead qualification intern, accounts payable intern.

• What information sources must the intern have access to?

  • E.g., social media, contact form responses, invoice database.

• What systems and databases should the intern fill in?

  • E.g., content calendar, CRM, accounts payable database.

• What is the specific sequence of tasks that the intern should be responsible for? For example, for a sales lead qualification intern:

  • Clean up the contact form data.

  • Query existing systems/databases and populate the lead record with all the identifiers that this record is associated with. If unclear, send an email for clarification.

  • Query external systems to check or complete information provided by the lead (e.g., job title, seniority, financials).

  • If the lead has high potential, notify an account manager immediately.

  • If the lead has unclear potential, schedule a 15-minute call with an SDR.

  • On a daily basis, review pending leads and remind salespeople of the 3 contacts they need to follow up with.

• What performance metrics should be used to evaluate if the intern is doing a good job?

  • E.g., number of tasks pending for human intervention, percentage of leads misclassified.

How to implement

GenAI should initially be implemented with minimal changes to internal processes and databases. It should stand on top of legacy platforms before deciding if legacy platforms must be migrated.

Here is a framework:

Takeaway messages

When evaluating GenAI contributions to the business, get your team to think about how they’d use an intern rather than how they’d use an AGI, to avoid falling into science fiction scenarios. Aim to augment and accelerate your existing business processes before reinventing how your customers and employees interact with your company.

This does not mean that businesses should disregard moonshot ideas. But they need to develop AI fitness before going for the Olympics.