AI is amazing, but it's not magic!

It's easy to assume AI agents are magic, but the reality is there are still some fundamental limitations to the technology.  Getting the most out of AI agents like those in DoubleO means understanding what you can, and can't do.

‍

What CAN'T you do with AI agents in DoubleO?

Two of the biggest limitations with AI are 1) how much "context" any given AI action can intake, and 2) how well AI can "build upon" previous work. These limitations are core to the technology, and not specific to DoubleO. For the most part, we've built in tools to deal with these limitations for you, so you shouldn't need to worry about them in most cases. However, there are two very common use cases people assume AI should be great at, which are not doable using the current technology:

1. You cannot use DoubleO to "build an exhaustive list"
   1. For example: you can't "find all Mexican restaurants in the US" or "find all the tennis coaches in Los Angeles".
   2. You CAN however find shorter lists (e.g. "the top 10 restaurants" or "a list of common SaaS tools") or bring your own list and take action on each item in the list (such as enriching each lead).
2. You cannot use DoubleO to "analyze massive amounts of data all in one step"
   1. For example: you can't ask an agent to "analyze all my sales call transcripts and identify themes" or "analyze all the data you can find about the Food and Beverage market and identify trends"
   2. You CAN however deeply analyze lots of data in smaller sub-steps, such as analyzing each call 1-by-1 then analyzing the summaries, or researching first the competitive landscape then the market size then recent trends.

‍

Do's and Don'ts for Using DoubleO

A few things to keep in mind, as you build workflows and give instructions to agents:

‍

Describe what should happen step-by-step as if you were doing it manually

✅ GOOD: "Every day review my calendar, then for each meeting with external attendees, find and review the attendee's LinkedIn to find their current role and priorities, then check our CRM for recent interactions to identify any action items or goals, then email me pre-call sales notes."

❌ BAD:  "Research all the people I'm meeting with today and give me pre-call sales notes."

‍

Describe what to do, not how to do it.

✅ GOOD: "Compare spreadsheet 1 and spreadsheet 2 and list any discrepancies."

❌ BAD:  "Search column A in spreadsheet 1, then search column B in spreadsheet 2, consider if the values are the same, then conclude if there are any discrepancies."

‍

Be specific about where to find data sources.

✅ GOOD: "Review the company's investor page and the google sheet at {{URL}} to find the company's financials."

❌ BAD:  "Find the company's financials."

‍

Research deeply, but not broadly

✅ GOOD: "Find me all the info about company X."

❌ BAD:  "Find me all the companies that fit X criteria."

‍

If an integration isn't supported, use an API call

✅ GOOD: "Get the support tickets from {{obscure support software}}."

❌ BAD:  "Get the support tickets from {{obscure support software}} by using a custom API call."

‍

If you ever have trouble, reach out to support@doubleo.ai!