when @biilmann.blog first started talking about AX like UX and DX, I had philological objections but we definitely need a term for how you get other systems and services ready so agentic AI can work with them effectively, including stopping agents using up too many resources by trying over and again

as I've suggested before, when there's an AI budget around, using some of it for things that help the humans and do the IT housekeeping as well doing what @charlestbetz.bsky.social reminds me is needed to make the AI work well, will pay dividends however the AI side of things works out long term

I asked @biilmann.blog how to build good AX; @daveon.bsky.social tells me how to manage & monitor MCP, @cote.io and @knime.com bring up data architecture, @amandaksilver.bsky.social talks me through how to set up agentic AI to pay attention to your AX work - and how that's working out at Microsoft

especially if you're one of the 30% of IDC clients working to modernise your software portfolio to work better with AI, APIs are key - and not always well implemented: @seanblanchfield.bsky.social tells me about the problems with common APIs that can derail agents and the Jentic tool to spot that

the thing I was fascinated (and should not have been surprised) to hear from @daveon.bsky.social about how malleable and self-updating MCP tools are: which is nice when it means they get better and not at all nice when it means they stop working or spontaneously switch you to a different vendor