Straight answers

Because the hard part isn’t the code. The model is the easy part. What makes AI give right answers is your knowledge written down once, your numbers defined and agreed, an owner who approves every change, and upkeep every month. That’s our method, with tested skills from a Hub in daily use across sales, marketing and support.

We work with your CTO and technology team, not around them. If you have a strong platform team, we can build the Hub into your own platform: your engineers build and own it, and we supply the method, the skills, the specs and the checks. Either way, your engineers stay on your product.

Yes, and we want you to. Your library has its own place for your skills, and small apps can run as pages in your Claude. We train your team as we go, and hand over is part of the plan: we’re finished when someone on your team has shipped a release themselves. If you’d rather we built something, we scope it and price it in days.

You own all of it. Everything runs in your own accounts: your GitHub, your Claude, your cloud. Your knowledge is in plain files that come with you if you leave, and Kixi is a guest you can remove in one click. You can take over any time after the foundation is built, once we’ve trained your maintainer.

A Hub super admin who’s responsible for the whole Hub, and an owner for each area of knowledge who approves its changes. They all get admin training at go-live. Only one technical person, ours or yours, works in GitHub. Everyone else suggests and approves changes in the browser.

Yes, and it’s included. Every feature comes with a 60 to 90 minute training session, a guide for your champions, and 30 days of tuning after it goes live. Admins and owners get their own training at go-live.

Every business starts in a different place, so we don’t sell a boxed price list. Each phase is a fixed fee, worked out from an estimate in days, and each project is scoped before it’s priced.

Book a demo and let’s discuss your needs and where you are on your AI journey.

Most of your Hub runs on your people’s Claude seats. On Claude’s Team plan, that’s a set price per person each month, not a bill that grows with use. Claude has controls built in, too:

• Higher-usage seats for power users. Anthropic offers seats with five times the usage of a standard seat, or more, at a set monthly price: far cheaper for your heaviest users than paying for the same work by use.
• Limits on extra usage. Your admins decide whether people can go past their seat’s allowance, and can set spend limits for each person and for the whole organisation.

Where answers are billed by use, such as in Slack or on your portal for people without a seat, you set the spend limits there too. And our reports fetch the data first and then do the maths, rather than asking the AI to read everything, so they stay quick and cheap.

See Claude’s plans and prices

Some, and you’ll know them up front. Your people’s Claude seats are paid to Anthropic, not to us. Hosting is a small monthly cost if you have a portal. Support from Kixi is your choice: we can look after it and keep it up to date, keep a Kixi principal alongside to lead what’s next, or hand it over to your team, with nothing more to pay us.

Book a demo and let’s discuss your needs and where you are on your AI journey.

No. Claude’s business plans and API don’t train on your data by default. Your knowledge lives in your own private GitHub, and Claude reads your systems as the person asking, with their own permissions.

Your library is in your own GitHub, and a portal, if you have one, runs in your own cloud (Google Cloud, Azure or Cloudflare). Claude’s processing is in the US or global. If any of your data has to stay in New Zealand or Australia, tell us in the Systems and data review and we’ll design around it.

In Claude, each person sees only what their own logins allow. With a Hub portal, roles come from your Google or Microsoft groups, each with a named owner, and access is reviewed every quarter. Knowledge for one group, such as your leadership team, can sit in its own restricted area, tested for leaks on every release.

Only when a person confirms it, such as a click on Create lead, and only where you agree. Those confirmations can’t be switched off.

The ones your teams already use: your CRM (such as HubSpot or Salesforce), your helpdesk (such as Zendesk or Intercom), Google Drive or SharePoint, Slack or Teams, Xero, and your data warehouse. It reads them live, as the person asking. For your own systems we can build a connector.

AI without your knowledge often does. Anthropic’s own team found about 21% of answers right without written knowledge, and over 95% with it. Every page in your library has an owner and a review date, your top questions are re-tested every month, and the Hub is tested to say “I don’t know” when it doesn’t. When someone spots a wrong answer, one click sends a fix to the page’s owner.

We test for it. Each model is retired with at least 60 days’ notice, about once a year, and your top questions are asked again on each new model. Your knowledge is plain files in an open format, so it isn’t locked to Claude.

Yes. They can ask in Slack, or on your Hub portal, with spend limits you set.

They do different jobs and can run side by side. Copilot reads everything in SharePoint you’re allowed to open, including old copies and drafts. Your AI Hub is a small, owned set of knowledge that decides prices, quotes and what your numbers mean, with an owner, an approval and a review date on every page.