Learning infrastructure to scale AI
Your capability strategy is blocked by bad data
Every workforce capability plan rests on your skills data and learning content. But half of enterprise content has no usable description, and your skills data is chaotic. Your people waste time finding the right resources and answers.
Filtered Intelligence breaks your content into parts, describes and quality scores every one, and ties them to skills and roles data that is authoritative.
We are the essential layer for enterprises focused on:
- AI readiness
- Workforce transformation
- Optimising AI investments
Trusted by NatWest · GSK · Swisscom · Reaches your agents over MCP
✓The skills to work on first
✓The exact content to use
✓Why it was picked
Built for this personA manager asks a question. The answer comes back shaped around them.
Not a list of courses.
External providers
O'ReillygetAbstractLearning platforms
DegreedWorkday LearnInternal content
SCORM packagesRecorded sessionsPeople records
RolesSkillsDelivery surface
CopilotTeamsThe answer is only as good as what the AI can reach.
We describe every part.
Your content becomes context.
It knows the role and the level.
Leading Others: Setting Direction
Best match for the skill.
Leadership Essentials for New Managers
A near duplicate, so it waits its turn.
Anything AI serves has to be worth the time.
Relevance, similarity and duplication decide which one shows up.
First-Time Leader Toolkit
45 minutes · built in house
Nobody has 45 minutes. They have the four they need.
Every part stands on its own.
One view of what your people can do.
AI skills are where the next money goes.
Your knowledge → Your context → The answer
The same answer, with the working shown.
Every pick shows its reasoning.
Saved against manual skills mappingGSK
Faster content search and curationVerified across deployments
Trusted by leading enterprises
of learning content has no usable description.
An agent chooses what to serve by reading descriptions and tags. No AI strategy works on foundations like that, which is why Filtered starts here.
- AI cannot read your learning content
- Your skills data sits apart from it
- Obsolescence and duplication go unseen
- Your platforms only see part of it
Example
One question. Two very different answers.
A team leader opens Copilot the week before a new manager starts. The answer depends on whether anything underneath can reach your learning content.
Your stack today
Your own team wrote a toolkit for exactly this. The agent cannot see it. The manager opens a browser tab instead.
With Filtered underneath
Same agent, same question, same content. The difference is the layer underneath.
AI workforce evolution
AI transformation stalls on the data underneath it
Four things have to be true before AI changes how work gets done. Each one depends on knowing what your people can do. Most firms do not have that data yet.
Strategic alignment
AI work owned at the top and tied to business goals, rather than run off to one side as a tech project.
Where we come inWe show your skills coverage against those goals. You see which have real strength behind them and which have almost none.
Capability mapping
A real measure of readiness. Not just who can use the tools, but who shows good judgment and who can lead the change.
Where we come inWe build the frame it hangs on. Every part and every record ties to a skill, a role, an opening and a level. A gap then has something attached to it.
Embedded upskilling
Learning that happens in the flow of work, in small pieces, so skills keep up with the tools.
Where we come inWe make the pieces. A 45-minute course becomes parts that stand alone, served inside Copilot and Teams at the moment someone needs them.
Role redesign
Roles rewritten around the human work that matters most, once AI takes on the rest.
Where we come inWe hold the skills and role data a redesign runs on. When a role changes, you see the skills it now needs, what covers them, and who is already close.
Filtered Intelligence
How it works
Four parts. Together they turn your learning content and skills data into something your AI can use.
Ingest Content
The useful five minutes is buried in a file that runs for forty-five. Half of it has no usable description. We break every course and document into parts, write a description for each one, and keep them all in one place.
Map Skills
You cannot see who can do what. We read your learning content and your skills records together, so every part connects to a skill, a role, an opening and a level.
Signal Quality
Obsolescence and duplication go unnoticed. We score every part continuously, so the best one surfaces and the rest stops getting in the way.
Connect AI
Your assistant answers from the internet because it cannot reach your content. We open the layer to any agent over MCP, so it answers from your own content. Copilot, Teams, Claude and ChatGPT Enterprise work on day one.
Architecture
How Filtered fits into your systems
Filtered is a layer, not a place people go. Your learning content and skills data rise into it, structure comes back out, and your people keep using the tools they already open.
Where your content and job data already sits
Filtered Intelligence
Where your people already work
Platform integrations
Works inside the systems you already run
Filtered connects by API to what you have, including your HRIS, LMS, LXP and content suppliers. Nothing gets replaced.
Where Filtered fits
Alongside or as an alternative to your existing stack
Most teams start by making what they already own work harder. Some use Filtered instead of buying a platform. Both work.
| Capability | Your LXP with Filtered underneath | Your LXP on its own |
|---|---|---|
| Level of detail | ✓Every part inside a file, findable on its own | ~Whole courses and catalogue records |
| Reach across the estate | ✓All your learning content, wherever it sits | ~What has been loaded into it |
| Skills, roles and openings | ✓Read from your material and your people records together | ~Connects to your other systems, with the tagging kept up by your team |
| Quality | ✓Every part scored for relevance and duplication, continuously | ~Ratings and use, rather than a score per part |
| What an agent retrieves over MCP | ✓Everything, whatever system it sits in | ~Its own catalogue and its own records |
| Capability | Filtered Intelligence | Traditional LXP |
|---|---|---|
| New employee portal | ✓None. Works in the tools they use | ✕Another place to log in |
| What an agent retrieves over MCP | ✓Every connected system, part by part | ~Its own catalogue, where MCP is offered |
| Where your content is processed | ✓In your own environment | ~Depends on the supplier |
| Typical yearly cost | ✓From £70K | ✕£500K to £1.5M+ |
| Skills gaps | ✓Kept up to date across every source | ~Against the list the platform holds |
| Capability | Filtered Intelligence | Building in-house |
|---|---|---|
| Time to go live | ✓Weeks | ✕12 to 18 months, typically |
| Who keeps it running | ✓We do | ✕Your own engineers |
| MCP | ✓Working on day one | ~You build it yourself |
| Real cost over three years | ✓A licence fee you can plan for | ✕Engineers, servers and upkeep |
| Tagging to your framework | ✓Done for you by Map Skills | ✕Built by hand |
Customer story
ECITB turned weeks of work into seconds
If you do not know what Filtered are doing, you need to go and understand it.Steve RickHead of Digital Learning, ECITB
Building a learning pathway or sourcing the right content consumed many hours of reading, sorting and tagging.
Filtered's API surfaces the right content in seconds, with pathways that build themselves.
Over £1m saved a year from digital transformation, with the LXP scaled to thousands of users.
Questions
Typical customer queries
What are the four parts of Filtered Intelligence?
Ingest Content takes in your learning content, whoever made it, and breaks it into parts. Map Skills ties those parts to your skills, roles and open positions. Signal Quality scores each part for relevance, obsolescence and duplication. Connect AI opens the whole layer to any agent over MCP.
We already have an LXP. Does Filtered replace it?
Only if you want it to. Filtered is a layer, not a portal. Most teams put it underneath the platform they own, so that platform can see every part rather than whole courses, and can read your people records rather than a list someone uploads by hand. Some teams look at the results and decide they no longer need the portal. That is a choice, not a condition.
Why can we not just point AI at the content we already have?
Because about half of it has no usable description. An agent chooses what to serve by reading descriptions and tags. Where those are missing, wrong, or written to sell a course rather than describe it, the agent either misses the content or serves the wrong thing. Filtered writes a usable description for every part first. That is what the rest is built on.
What is MCP, and why does it matter?
The Model Context Protocol is an open standard. It lets an AI agent find and use data that sits outside itself. Filtered opens your layer through MCP, so any agent that speaks it can reach your material without a custom build. What your people made stops being invisible.
Is our material safe? Does it leave our systems?
Your content stays inside your boundary. We run our own models rather than sending your material to a public one. Nothing you hold is used to train anyone else.
What does Filtered connect to?
Filtered connects by API to the tools you already run, including Workday, Microsoft, Oracle, SAP, Google and Okta. Nothing gets replaced.
Get started
Your AI needs something underneath it
See how Filtered joins up your learning content and skills data, in a walkthrough set around your own systems.