Squid DataMind • The enterprise AI context platform

One trusted business context for every agent.Traceable. Governed. Fast.

DataMind connects to your databases, apps and documents where they live, and works out what your data means. Each agent gets just the context it needs, within your permissions and traced to its source. So agents give more accurate answers, safely, faster and at lower cost. No migration, and no lock-in on your context.

See it on your dataHow it works
Trusted by data and AI teams at some of the world's largest enterprises
NTTNECChicago BullsPritzker GroupTevaAlpineMonogotoNVIDIA
Why context matters

The model isn’t the problem. Missing context is.

Without a shared business context, agents give confident wrong answers, show people data they shouldn’t see, and cost more to run. Here’s why.

  • GROSS MARGINSQL34.8%BI30.5%Finance32.1%32.1% ✓approved

    Metric drift

    One KPI, several definitions across SQL, BI and finance. An agent has to use the approved one, and show it.

  • CUSTOMERCRMACC-2291ERPV-88104SupportJ. SmithJane Smith ✓one customer

    Identity

    The same customer sits in many systems, under different IDs. An agent has to know it’s one customer.

  • CUSTOMER TIERCRMGoldBillingSilverGold ✓CRM wins,reason logged

    Disagreement

    Systems hold different truths about the same thing. An agent has to know which one wins, and why.

  • PRODUCT IDBeforeprod_idAfterproduct_codeProduct ✓still mapped

    Constant change

    Schemas keep changing across sources. An agent’s context has to stay mapped, without a rebuild.

  • REVENUE BY REGIONAsked by an EMEA managerEMEA$4.2M ✓APAChiddenUShidden

    Permissions

    People are cleared to see different data. An agent has to show each person only what they may see.

  • CONTEXT PER QUESTIONWithouteverything, pieced togetherslow · $$$$Withjust what it needsfast · $ ✓

    Speed and cost

    Piecing data together for every question is slow and costly. An agent has to get just the context it needs.

How DataMind works

Your business context. Automatically built. Rapidly confirmed. Safely served.

DataMind sits between your systems and every agent. It reaches your data where it lives, works out what it means for your experts to confirm, and serves it live, with your rules and tests on every layer.

Tap the + on any part of the diagram to see how it works.

Your data · in place
  • Databases and warehousesPostgreSQL · Oracle · MongoDB · Snowflake · Databricks
  • Business apps and serviceSalesforce · SAP · HubSpot · ServiceNow · Zendesk · Jira
  • Documents and APIsSharePoint · Slack · PDF · REST · email · any MCP server

100+ connectors · no migration

Govern · test · secure · every call
Squid DataMind
Understand your businessGenerated from your data · confirmed by your experts
  • Meaning
    • Semantic layer
    • Business taxonomy
  • Connections
    • Knowledge graph
    • Ontology
  • Origin
    • Context graph
    • Lineage
  • Reach
    • Federated query
    • Memory and cache
Business context
Answer every questionLive · checked · cited
  1. Know who’s asking
  2. Get live data
  3. Act, under control
Any agent · open protocols
  • Your agentsClaude · ChatGPT · Gemini · Copilot Studio · your own
  • Squid’s agentsOntology · Lineage · Service · Sales · Finance
  • Where teams use itSquid Studio · Slack · Teams · Helpdesk · CRM · console

MCP · A2A · REST API · SDKs

Runs where you choose
  • Squid cloud
  • Your VPC · AWS · Azure · Google Cloud
  • On-prem
  • Air-gapped

Regional data residency

  • Squid DataMind
  • Yours: agents, data, infrastructure
  • Govern, test and secure: on every call
  • Purple arrows: what agents receive
Inside DataMind

A closer look at DataMind’s core components.

Each component from the diagram above: what it does and how it looks in the product.

Meaning

Semantic layer

One definition of every metric, mapped automatically across all your data stores.

Ask three teams for “gross margin” and you may get three numbers. A semantic layer records the one agreed definition, so every report and every agent calculates it the same way.

The basics, covered

  • Shared metric definitions, in one place

What DataMind adds

  • Understands how the same business data is stored in different systems, and maps it for you
  • Works with the semantic layer and catalog you already have
  • Drafted for you from your schemas and how the data is actually used
  • Owners approve, and every change is versioned
  • Served live to any agent
Meaning

Business taxonomy

Your business terms, how they group, and what else people call them.

Every company has its own words and its own way of grouping things: product family → product → plan, or region → market. A business taxonomy records them, and which words mean the same thing, so “rebate” and “allowance” lead to the same data.

The basics, covered

  • Business terms and their definitions

What DataMind adds

  • How terms group, from product family down to plan
  • Synonyms included, so people can ask in their own words
  • Applied to every answer
Connections

Knowledge graph

The same customer, order or contract, recognised in every system.

One customer can be an account in your CRM, a vendor ID in your ERP and a name in a support ticket. A knowledge graph recognises them as the same customer and keeps track of how customers, orders and contracts relate.

The basics, covered

  • Customers, orders and contracts linked across systems

What DataMind adds

  • Documents and tickets linked too, not only tables
  • Answers cite documents down to the page and clause, next to the database rows
Connections

Ontology

A live model of your business: industry model, business objects and rules.

An ontology is the map of your business: what kinds of things exist (customers, contracts, plants), how they relate, and the rules they follow. DataMind starts from an industry model and adds your own business objects and rules, so you don’t build it by hand. Every suggested link comes with a confidence score and its reasoning: the inference states its own uncertainty, and your experts make the call.

The basics, covered

  • A model of your business objects and the rules between them

What DataMind adds

  • Starts from an industry model, not a blank page
  • Extended automatically from your schemas, code and documents
  • Every link scored, with written reasoning; your experts confirm it
  • Kept current in real time
Origin

Context graph

Decisions that compound.

Data shows what happened. A context graph also keeps why: the decisions, approvals and exceptions behind it. The next person or agent starts from what was already learned.

The basics, covered

  • A record of past decisions and approvals

What DataMind adds

  • Decision traces and approved changes, kept for the next agent
  • Linked to the records they touched
Origin

Lineage

Every number traced to its source row.

Lineage shows where a number came from: which systems, tables and calculations it passed through. When a figure looks wrong, you can follow it back.

The basics, covered

  • Field-level lineage across systems

What DataMind adds

  • Calculation checks back to the original rows
  • Every answer traces back to its sources
Reach

Federated query engine

Your data, queried where it lives.

Instead of copying all your data into one place first, a federated engine sends each question to the systems where the data already lives and combines the results.

The basics, covered

  • Queries across many sources at once, with no migration
  • No pipelines, no copies

What DataMind adds

  • Native dialect for each source; it corrects its own failed queries
  • Runs in place across 12 database engines
  • Review the SQL before it runs, and inspect every query that ran
  • Read-only: generated queries that would write are rejected
Reach

Memory and cache

Faster and cheaper every time.

Agents shouldn’t start from zero on every question. Short-term memory keeps track of the task at hand. Long-term memory keeps what has been learned and confirmed, for every agent. A cache reuses answers and queries, so agents respond faster, cost less to run and put less load on your systems.

The basics, covered

  • Memory within a conversation

What DataMind adds

  • Long-term memory of what’s confirmed, shared across agents
  • A cache of answers and queries, for faster answers and lower AI costs
  • Less load on your source systems
Govern every call · Secure and comply

Governance

Your rules on every call. Your data stays yours.

Governance decides who can see what, and what agents are allowed to do. With DataMind, a person sees only what they can already see in the source system, and every action is logged.

The basics, covered

  • Encryption (AES-256, TLS 1.2+) and a full audit log
  • SOC 2 Type II, ISO 27001, HIPAA, GDPR; no training on your data

What DataMind adds

  • Permissions from your identity provider (Okta, Entra ID, Auth0 and more), checked at query time, never in a prompt
  • Approval before any write-back, with rollback
  • Prompt screening, tool scope and rate limits
Test and monitor

Testing and monitoring

Before launch and after.

Before an agent goes live, you want to know it gets the right answers. After launch, you want to know if that changes. Testing checks answers against ones you know are right; monitoring watches accuracy, speed and cost in production.

The basics, covered

  • Monitoring and alerts for accuracy, speed and cost

What DataMind adds

  • Tested against your own known-good answers before launch
  • Quality checks on responses, SQL and extractions
  • Replayable traces from question to action
Customers

Value at enterprise scale, delivered now.

Leading companies, from telecom to financial services and healthcare, rely on Squid for the business context that makes their AI agents successful.

1.5M

Support tickets a year, with about 5 minutes saved on each: roughly 60 people’s worth of capacity.

Cloud communications provider · Telecom and technology

10M+

Claims across 5 live systems, with every answer cited.

Healthcare payer · Health and life sciences

100%

Of figures traced to the query behind them, in agents a model-risk team can sign off.

Financial services

80%

Faster issue resolution, with half the escalations.

Global connectivity provider

95% less build effort for each agent you add, because every new agent reuses the same context.

NTTNECChicago BullsPritzker Group
TevaAlpineMonogotoNVIDIA
How to start

Start with one use case.

Prove it on your own schemas and data before you commit.

  1. Pick one use case

    The one your business needs most.

    • Customer service: tickets, manuals, policies
    • Sales: pricing and deal terms
    • Finance: revenue, cost and margin
    • Data team: explain a KPI move, reconcile a metric

  2. We design it with you

    In Squid’s cloud, from your schemas and sample or synthetic data: the business context, the access rules and how agents will use it. No customer data needed yet.

  3. You prove it in yours

    DataMind runs in your own cloud on realistic data that never leaves it. Agents start by recommending: they find and explain, your people act. Every test passes or fails with evidence, in a written report.

See it on your data

Give every agent one trusted business context.

Traceable. Governed. Fast. On your data, with no migration.

See it on your data