Choose Your Approach

Two decisions, in order. First, whether a chatbot is the right tool at all. Then, which kind to build.

Should this be a chatbot?

Signs it’s a good fit

  • People ask your team the same questions over and over
  • The answers already exist in writing
  • Somebody owns that content and can keep it current
  • The questions are informational — “when is the deadline,” not “please change my registration”
  • Your audience is willing to self-serve

Signs it isn’t

  • The answers live in people’s heads, not in documents
  • Your content is out of date, scattered, or contradicts itself
  • The real problem is that your website is hard to navigate
  • The volume is low — a handful of questions a month
  • The answers have to be exact for legal or financial reasons
  • Nobody can commit to maintaining it
A chatbot built on bad content doesn’t fail quietly. It produces confident, wrong answers faster than your website ever could.

Things that often work better

If the fit isn’t there, one of these usually solves the actual problem:

  • Fix your site search. Often the whole issue.
  • Rewrite the FAQ. If the answers aren’t clear on the page, they won’t be clear in a chatbot either.
  • Build a form for anything transactional.
  • Fix your information architecture. If people can’t find things, a chatbot is a workaround, not a fix.
  • Use saved replies in a shared inbox. Low effort, immediate.

These are real solutions, not consolation prizes. Coming to us for a chatbot and leaving with a clearer website is a good outcome, and a common one.

Gem or managed chatbot?

  Build it yourself Build it with us
Who builds it You AI Enablement, with you
How long An afternoon Weeks to months
Who can use it ND people with a Google account Anyone, including the public
Where it lives Inside Gemini Embedded on your site, or its own page
IT involvement None Some, for deployment
Cost Covered by existing licensing Usage-based; your unit pays
Content Files you point it at in Drive Connected sources, larger sets
Usage data Minimal Available to review
Data rules Identical Identical

That last row is not a formatting error. Building a Gem privately doesn’t change what data you’re allowed to use.

Choose a Gem when

  • Your audience is internal
  • Your content fits in a handful of documents
  • You want something working this week
  • You’re testing whether the idea holds up
  • You’re not sure yet

Choose a managed chatbot when

  • Your audience includes people without ND accounts
  • Volume is high and sustained
  • It needs to live on your public website
  • You need to see how it’s being used
  • Your content is large and changes often

Start with a Gem either way

Even if you’re confident you need a managed chatbot, build a Gem first.

It’s the cheapest way to find out whether your content can actually answer your users’ questions. If the Gem works, you have a working prototype and a proven set of content. If it doesn’t, you learned in an afternoon what would otherwise have taken two months to discover.

Before you ask us for a managed chatbot

You’ll need these in place. Not paperwork — these are the things that determine whether it works.

  • A sponsor with authority to make decisions
  • A content owner accountable for accuracy, now and later
  • A budget owner and a FOAPAL
  • A clearly defined audience
  • Content that already exists in writing
  • People willing to test it properly