Enterprise AI has an accuracy problem, and it isn’t the models.
Ask a good LLM to calculate net revenue retention and you’ll get a fast, fluent, confident answer. Ask again next quarter and you may get a different one, because the model is deriving the logic from scratch every time. Your analytics team settled that definition years ago, wrote it into a workflow, and had it reviewed. The model has never seen it.
That gap, between what a model can infer on its own and what your business has already agreed on, is where most enterprise AI pilots stall.
The latest AI capabilities in Alteryx One are built to solve this challenge. The theme of our 2026 Summer Release is AI with ROI, focused on three areas where Alteryx improves the economics of AI.
- Answers are more trustworthy. Every response traces back to the same governed workflows your teams already maintain, keeping answers consistent across users and AI tools. NextWave recently demonstrated a 100% accuracy rate using Alteryx in a complex reconciliation process, catching all 11 breaks, compared to a general-purpose LLM that caught 9 but missed 2 completely.
- Token spend goes down. Complex analytics run inside Alteryx instead of being re-derived by a model on every question. The result is predictable, governed compute in place of token spend that scales unpredictably. In the same reconciliation use case, NextWave reduced LLM token consumption by 20x using Alteryx to execute the established logic rather than asking the model to reason through the entire process each time.
- Outcomes come faster. Agents reuse workflows your team has already built rather than rediscovering business logic from zero. Using Alteryx One, Big Query, and Gemini, Papa Johns reduced franchise billing and reconciliation processes from 300+ hours to 2 hours.
The governed foundation, now on every AI surface
We describe that foundation as VURA: visible, understandable, repeatable, and auditable. An answer should show its inputs, explain its steps, produce the same result twice, and leave a trail someone can review.
This release extends VURA to the AI surfaces your business already uses. Ask a question in Claude, ChatGPT, or Alteryx One, and it resolves against the same approved logic your analytics team maintains.
New ways to put trusted logic to work
The capabilities in this release extend the same governed foundation across a natural progression: ask a question, create a reusable workflow, turn that logic into an agent, and make it available wherever your teams work.
- Ask Alteryx: Ask a business question in plain language, inside Alteryx One. Ask Alteryx delivers a trusted answer when one already exists, and builds a new, reusable workflow when it doesn’t. Users can inspect, edit, and verify every output, or summarize the workflow, so a quick answer today becomes a governed asset tomorrow.
- Agent Studio: Turn any approved dataset into a conversational agent in a few clicks, with no rebuilding of the underlying logic. Finance, operations, and other teams get an agent they can simply ask. Analytics teams keep full control of the data and logic behind every response. Agent Studio is a built-in Alteryx One capability with no additional token costs each time an agent runs. Building and using Alteryx Agents in Agent Studio is in public preview.
- Alteryx MCP Server and Skills: Connect Alteryx to whatever AI platform you already work in. Your agent interacts with Alteryx the way a person would: finding the right data, building a multi-step workflow, and turning it into something governed and repeatable. Alteryx Skills teaches AI coding tools like Codex, Copilot, Claude Code, Gemini CLI, and others, to build Alteryx workflows using the same patterns Ask Alteryx uses.
- Alteryx Insights for OpenAI: Available now through the ChatGPT Plugin Directory, this feature brings analyst-approved data, calculations, and workflows into ChatGPT. A business user can investigate a revenue variance or resolve a reconciliation question without opening Alteryx or holding a seat. Support for Claude, Gemini, Slack, and Microsoft Teams is on the way.
Why not just point the model at my data warehouse?
Fair question, and you should — for retrieval. Databricks and Snowflake now do real analytics too, not just storage.
What they’re not built for is letting the people who hold the business rules encode them directly. An FP&A analyst knows that revenue excludes intercompany transfers, EMEA follows a different fiscal calendar, and three SKUs were reclassified in March.
With Alteryx, the analyst can translate that knowledge into reusable logic without relying on someone else to interpret it. In Databricks or Snowflake, the same rule usually means a ticket to a data engineering team who has to translate it into code first.
Alteryx sits between the agent and logic. The model keeps doing what it’s good at — understanding the question and handling the conversation. The calculation runs against business rules a human already approved.
Your AI tools like Claude, ChatGPT, Gemini, and Copilot also become more useful when the answers they surface are grounded in logic your business owns. Your existing stack stays in place and gains a trusted business logic layer.
Putting AI to work with confidence
The next phase of enterprise AI will be defined by more than model capability. It will be defined by how well organizations connect those models to the context, logic, and accountability that make their businesses run.
That is what the Alteryx 2026 Summer Release delivers, with a governed foundation that reaches from the analyst’s workflow to the business user’s conversation and into enterprise agentic systems. AI that can show its work, repeat it reliably, and deliver value at scale. AI with ROI.
I encourage you to explore how the new capabilities deliver value to business and IT teams here and dive into details in the release notes. To support and accelerate our users’ journey into agentic analytics, Alteryx is also offering a new customer learning path for current users.