Meaning before models
A three-layer ontology resolves your vocabulary before a single token is generated. The system knows a claim in freight isn't a claim in insurance.
THE PLATFORM
One environment for the entire life of an enterprise AI solution.
flexday.ai is where a business question becomes a governed, monitored application, with an ontology underneath it, connectors behind it and an audit trail attached to every answer it gives.
platform.flexday.ai
PLATFORM LAYERS
Experience
consoles · queues · chat · worksheets · launchpads
Orchestration
agents · flows · tools · HITL gates · schedules
Reasoning engines
SpeciphicAsk · StructIQ · SpeciphicView
Meaning of Data
canon · derived · customer ontology
Data foundation
connectors · fact bases · data models · lineage
WHAT MAKES IT DIFFERENT
Six things that are hard to bolt on afterwards.
Anything can generate an agent now. These are the parts that decide whether the agent survives contact with a real enterprise.
01
02
Applications, not just answers
The builder generates working interfaces, review queues, approval flows, operator consoles, worksheets, because most enterprise work isn't a conversation.
03
Works before your data is ready
No warehouse? No clean master data? The platform builds against documents, screens, exports and synthetic sets, then swaps in live systems without a rewrite.
04
Three altitudes of control
No-code, low-code and full code on the same solution. Business users and engineers work on one artifact instead of two products.
05
Accuracy as a measured property
Golden sets, regression runs, confidence thresholds and drift alerts are built into the release path, not added by a consultant later.
Trust and Security →06
Three engines, one context
Documents, structured data and video are reasoned over together. Most stacks make you pick one and integrate the rest yourself.
THE BUILD SURFACE
What a builder actually touches.
The vocabulary of the platform is deliberately small. Learn six objects and you can build anything on it.
OBJECT
Solutions
The deployable unit. A solution owns its agents, screens, data models, permissions and evaluation set, and it versions as a whole.
OBJECT
Agents
A scoped reasoner with tools, a role, guardrails and an ontology view. Agents call other agents; every hop is logged.
OBJECT
Flows
Deterministic orchestration around non-deterministic steps: branching, retries, schedules, approvals, compensation on failure.
OBJECT
Fact bases
Curated, permissioned corpora of documents and records with chunking, embedding and citation policy managed for you.
OBJECT
Data portraits
What the platform learned about a source: schema, semantics, joins, quality, freshness. This is Layer 2 of the ontology, made visible.
OBJECT
Launchpads
How a solution reaches people, embedded, standalone, in Teams, in an existing app, with role-based entitlements attached.
DEPLOYMENT
Runs where your data already lives.
flexday.ai deploys into your cloud tenancy or ours. Model choice is configurable per solution, including your own endpoints and regional constraints. Nothing about the architecture assumes we host your data.
- Your Azure or AWS tenancy, VNet/VPC-isolated, private endpoints, your keys
- Flexday-managed cloud, for speed of proof, with a documented exit path
- Hybrid, reasoning in your boundary, control plane managed
- Model flexibility, Claude by default; bring your own endpoint per solution
INTEROPERABILITY
It plays well with what you've bought.
You already have a warehouse, an iPaaS, a BI tool and probably a Copilot license.flexday.ai is designed to sit alongside those, not to replace them or to pretend they don't exist.
- Warehouses & lakes, Snowflake, Databricks, Fabric, BigQuery, Redshift
- Systems of record, SAP, Oracle, Salesforce, Workday, ServiceNow, AS400
- Content, SharePoint, Drive, S3, Confluence, network shares
- Downstream, Power BI, Tableau, Teams, Slack, your own front ends via API
PLATFORM SPECIFICATIONFLEXDAY.AI
- BUILD MODES
- No-code conversational builder · Low-code flow graph and expression editor · Full code source access with round-trip
- ONTOLOGY
- Three layers, canon (domain), derived (from your data), customer (your rules). OWL 2 / Turtle import and export.
- ENGINES
- SpeciphicAsk (documents & retrieval), StructIQ (structured analytics), SpeciphicView (image & video)
- INTERFACES
- Generated consoles, review queues, worksheets, approval flows, conversational surfaces, embeddable components, REST APIs
- GOVERNANCE
- RBAC from your IdP, row-level lineage, full prompt/context/decision audit, human-in-the-loop confidence gates, tenant isolation
- EVALUATION
- SME-owned golden sets, regression on every release, confidence calibration, drift and freshness alerting
- DEPLOYMENT
- Your Azure or AWS tenancy, Flexday-managed cloud, or hybrid. Private networking and customer-managed keys.
- COMPLIANCE
- SOC 2 · ISO 27001 · ISO 42001 · CCPA · USDP · HIPAA · GDPR · NIST CSF 2.0 · NIST 800-171
See it against your own data.
The fastest way to evaluate the platform is a scoped six-week POC on a process with real volume behind it.