WhAIQ · For integrators

When your agent needs a number, it asks WhAIQ

WhAIQ calculates what moves a company's numbers on the model of that company. Your agent asks the question and WhAIQ answers with the calculation: the same question gets the same answer, and the reasoning can be checked.

Il tuo agente chiede a WhAIQ il margine dopo un aumento di listino del 3%: via MCP o API WhAIQ restituisce il calcolo, +2,1 punti di margine e 74% di probabilità di centrare l'obiettivoYour agent asks WhAIQ for the margin after a 3% list-price increase: through MCP or API, WhAIQ returns the calculation, +2.1 margin points and a 74% probability of reaching the target
The problem

The agent runs the process. The number it gives is generated

What goes wrong today

  1. 01

    When a step needs a figure, like the margin after a price change, the language model writes it word by word.

  2. 02

    Ask the same question twice and the figure can change. Your client's controller notices.

  3. 03

    There is no model behind that figure to check, so someone ends up redoing the numbers by hand.

What your agent gets back

What comes back is a calculation

1

What moves what

The drivers of margin, cash and volumes on the client's KPI tree, and how much each one weighs.

2

The probability of reaching the target

For each choice, estimated by simulating the possible outcomes on the company's model.

3

Combinations compared

Several combinations of levers evaluated in the same calculation and ranked against the target.

Plug it in

Two ways in. One calculation

MCP

From your agents and assistants

Connect WhAIQ to the assistant or the agents you already use. Numbers calculated, not generated.

API

From your systems

Bring the calculation into ERP, CRM, orchestrators or wherever decisions already happen.

The model and the data

Whose model it is, and where it runs

Who builds it

Vedrai, you, or the client

Models are trained on the company's data, and the type of model changes with the use case. We can build it, you can build it as implementing partner, or the client's team can.

Where it runs

Servers in Europe

Model and data run on Vedrai servers in Europe, and Vedrai is ISO/IEC 27001 certified. Data can be hosted by us or read through connectors.

Who owns it

The client

The model stays the client's, and their data never improve models used by other clients.

Security and data
Questions integrators ask

Before you plug it in

How long does a calculation take?

Simulations run from milliseconds to minutes.

Who writes the result into my client's systems?

Your agent or your integration. WhAIQ returns the calculation, and you keep control of what gets written and where.

What kind of model does it use?

It depends on the use case: from predictive to causal models, through prescriptive models, optimization and our proprietary quantitative trees. All of them are trained on the company's data.

How do I try it?

Request partner test access: you connect WhAIQ through MCP to your agents and test it on a case from one of your clients.

Is there evidence that it makes agents better?

Yes. In YC-Bench, a public benchmark where an agent runs a start-up for a simulated year, all five models in the paired test did better with WhAI: 48 pairs out of 48. Read the research
Read next

WhAIQ for agent platforms

Enterprise suites and orchestrators

Read

WhAIQ for planning and BI suites

Probabilities and alternatives on the plan

Read

The Decision Layer

The research behind it

Read

Security and data

Europe, ISO/IEC 27001, the model stays the client's

Read

Try WhAIQ on a case from one of your clients

We give you test access and you connect it to your agents through MCP.