Home/Platform

More than talking to your data. An intelligence layer you can govern.

Optialy works live across the systems you already run, under your own definitions and permissions. Use it in our app, or inside Claude, Copilot, ChatGPT and Slack.

Start a pilotWatch the demo
What Optialy is

Governed. Cited. Computed.

Three properties every answer has, whichever assistant asked.

GovernedYour meaning, your access

Your field definitions, formulas and permissions apply to every answer. It inherits each system's access and never grants what a source denies.

CitedEvery figure traceable

Each number links to the record, system and time it came from, so the data can be separated from the interpretation.

ComputedCalculated, not estimated

Arithmetic runs as code outside the language model. The model plans the work; it does not make up the numbers.

Five ways teams try to close the gap today. Each one breaks.

×Move it to a warehouseEvery new question needs modelling first, and the copy is stale the moment it is made.
×Ask the data teamA queue measured in days, for a decision measured in hours.
×Buy each vendor's copilotAnswers stop at that vendor's boundary, and fail first on the custom fields where the business meaning lives.
×Export to a spreadsheetNo lineage, no dates, no citations. Nobody acts on a number they cannot check.
×Paste it into ChatGPTProbabilistic, weak at math, and blind to live private data.
Ask anywhere

One question. Four tools. The same answer.

Plug Optialy into the assistants your team already uses. Every one of them reaches your data through the same governed path.

Copilot
Claude
ChatGPT
Slack
Same number, same sources, same permissions, in every tool
How it works

Watch a question become an answer.

Six steps, every one visible. Agents plan and read; code does the math; nothing is acted on without a person.

Elicit 1

Asks before guessing

Why did our sales decline over the last two weeks?
Net sales or gross? → Net sales

Govern 2

Approved definitions, your access

“Net sales” = approved definition
Read as you, your access only
Same metric in every tool

Swarm 3

Agents read each source live, in parallel

Shopify
GA4
Klaviyo

Analyze 4

Math runs as code in a sandbox

# secure sandbox
d = join(shop, ga4, email)
d.decompose("net_sales")
→ 3 drivers found

Converge 5

One cited answer, with a chart

cited

Act 6

Drafts the fix; a person approves

Pause email flow BRestock alert to opsApprove
BeforeAn analyst exports from three tools, lines them up in a spreadsheet and replies days later.
With OptialyOne question, every relevant source read live, a cited answer and a drafted next step in minutes.

Scenario from our demo script, on an e-commerce environment with Shopify, GA4 and Klaviyo connected.

Measured, not claimed

Computed beats generated: 97.5% vs 65.8% of figures correct.

4,500 real marketing and commerce questions on live connected data, each graded blind by three AI judges against figures computed from the source systems, then checked manually by our team.

RAG pipeline0%
Optialy0%

Figures correct. Each square is 0.1% of graded figures.

Answer quality, out of 5

RAG pipeline0
Optialy0

Answers with a chart

RAG pipeline0
Optialy0

Why a RAG baseline

Plain ChatGPT or Claude can't connect to most of these systems, so many of the questions are impossible for them. The fair comparison is a pipeline that can.

Internal benchmark, September 2026. RAG pipeline: retrieval, then the language model reads the data and writes the figures itself (our earlier architecture).

Charts

Charts that build themselves, and stay live.

Every answer comes with an editable visualization. Keep it on a board and it refreshes itself.

Revenue and ad spend, weeklyShopify, Meta Ads
Revenue mixShopify
ROAS by campaignMeta, Google
Stock-out risk by weekInventory
Monday boardrefreshed just now
PowerPointWordPDFShare or fork the thread

Sample data

Principles

AI proposes. Code measures.

The rules every answer follows, whichever assistant asks.

Nothing is copied

Each system is read where the data lives, using the asking person's own access.

The smallest set of sources

Agents pick only the systems that can answer, and rule a source out rather than guess.

Every number from code

The AI plans the work. Calculation runs in a secure sandbox and can be recomputed.

Matched across systems

Customers, products and campaigns are matched across systems that name them differently.

Cited to the record

Each figure links to its record, system and retrieval time.

Asks before assuming

Clarifies what a word like "revenue" means once, then remembers it for your team.

One question. Every relevant source. Trusted wherever work happens.

Start with one question that costs you money today. We'll answer it on your live data.

Start a pilotWatch the product demo