As those who have followed our journey will know, we are prone to become extremely excited when it appears that AI will finally be in the hands of those who really need it - our business users who have endless ideas on how to transform their organisations with data without the time or tools to effect the change.
For years we have said the same thing: the people with the best ideas for transforming their organisations with data are almost never the people with the time or tools to build the solution. Business users can see the opportunity clearly. The backlog, the skills gap, and the fragility of hand-built automation get in the way.
That gap has finally closed. We believe our clients now have the right tools in their hands, and those tools are getting better every week.
How we see modern data automation - harness the limitless imagination of Claude with data automation that generates the same outputs with a set of inputs, every time.
What Changed
A handful of breakthroughs have landed at once, and together they have tipped the balance from interesting demo to production-ready:
Anthropic's Claude finally delivered on the original promise of generative AI with Opus 4.5 — accurate enough, and steerable enough, to trust with real work.
Alteryx One brought Auto Insights and a secure cloud portal to every client, out of the box.
MCP servers gave AI and data platforms a common language, so they can talk to each other without custom integration work.
Orchestra arrived to coordinate tasks across platforms, so you can run a process end to end without a data engineering team behind it.
None of these on its own would change much. Together, they remove the three things that used to make AI-driven data work fail: unreliable output, painful integration, and a dependency on scarce engineering skills.
The blueprint: generative, deterministic, generative
The mistake most often made is asking generative AI to do the whole job, including the part it is worst at.
We know large language models are brilliant at interpreting intent and generating a first draft, and a poor choice for a process that has to produce exactly the same result every time, such as a reconciliation or a regulatory return.
So we do not ask them to. Our blueprint is a sandwich:
Generative — use Claude once or twice to build an Alteryx workflow from the raw material you already have: meeting transcripts, specification documents, technical specs, schema, existing workflows, and API endpoints.
Deterministic — your team, or ours, fine-tunes, tests, and publishes that workflow in Alteryx. From that point it is fixed logic: same inputs, same outputs, every run, fully auditable.
Generative — point Claude back at the clean, trustworthy output to chat to the data, feed agents, and produce reports.
The deterministic middle is what allows a regulated business put this into production with confidence.
What we have built
We have spent this year building Claude IP on top of that blueprint. Today we can:
Create 90% complete Alteryx workflows from business meeting transcripts, specification documents, technical specs, and API endpoints — turning a scoping call into a working draft in minutes rather than days.
Connect Alteryx, Claude, and client systems over MCP for genuine end-to-end processes. One example we are proud of: generating a fund distribution from start to finish.
Unlock chat-to-data in seconds using Claude and Alteryx Auto Insights — the best answer we have seen to a request clients have been making for years, and with no semantic layer to build or maintain.
There is a security dividend, too. Because agents are fed only the specific data they need from the Alteryx output, you avoid exposing whole datasets and you cut the token burn that makes naive AI projects slow and expensive.
The result is speed. One of our larger clients expects a 5x increase in how quickly they develop and deploy workflows.
This clearly isn’t a marginal gain, it’s a step change!
So, what does this mean for you, and why are we the people to talk to?
Dan with the Alteryx Agentic AI team at Inspire Orlando, May 2026
Built with the people building the platform
At Inspire Orlando in May we were lucky enough to spend time with Jupiter and the Alteryx team building the MCP server elements of Alteryx One. We had been waiting years for something like this, and we were able to share exactly what a perfect version would look like for the kinds of problems our clients face.
In a matter of weeks they far exceeded what we had imagined. Alteryx One is now the easiest way we know for an organisation to put its data in front of agents to solve problems, not just display it in another spreadsheet or dashboard.
Where to start
If you are already using Claude or Copilot and want to do something more valuable than summarising documents, or you have existing workflows that need upgrading quickly, this is the moment to start.
The fastest way to see it is to watch it work on something of yours. Get in touch and we will run a free workshop for you.
We hope that you can see how enthused we are…to be continued !