Knowledge product types for educators: 2026 AI guide

If you earn £1M+ from your expertise, five knowledge product types are worth your time: micro-credentials, cohort programs, toolkits and templates, micro-courses, and on-demand research clinics.

Start with a micro-credential feeding a cohort. It packages your method as a credential, creates a repeatable revenue cycle, and produces the kind of structured data an AI agent can act on from day one.

• Micro-credentials: short, outcome-specific badges tied to your method
• Cohort programs: time-limited group runs with community and accountability built in
• Toolkits and templates: ready-to-use assets clients implement immediately
• Micro-courses: focused modules on one skill or decision, priced as entry points
• On-demand research clinics: advisory sessions triggered by a specific client problem

---

What is a knowledge product, and how does the N1 to N4 model work?

A knowledge product is a structured package of expertise that guides a decision or an implementation step. It is not raw content. A blog post is content. A playbook that walks a client through a specific decision, with rubrics and outputs, is a product.

The clearest planning model comes from the PAHO Virtual Campus classification of learning objects, which grades resources by granularity from atomic files up to full diploma programs.

Level	Name	What it is	Revenue model	Longevity
N1	Atomic object	Single file, prompt, template or rubric	One-off or bundle	Short term
N2	Lesson unit	One structured learning experience with an objective and an assessment	One-off or micro-bundle	Short term
N3	Modular course	Several lesson units on one topic, with outcomes and credentials	Cohort or subscription	Multi-year
N4	Program or diploma	Multiple courses forming a full pathway	Enterprise or cohort	Long term

Build at N1 first to test demand. Stack N1 objects into N2 units, then N3 courses, then N4 programs as the evidence grows. The mistake is starting at N4 because that is what the finished thing looks like in your head.

---

Which knowledge product type should you build?

Each type maps to a build effort band. A quick win takes under four weeks. A 90-day prototype takes a quarter. A multi-quarter build takes six months or more.

Product type	Best for	Effort to build	Scales with AI	Revenue model
Micro-course	Entry-level funnel, lead conversion	Quick win	High	One-off or bundle
Micro-credential	Outcome-first buyers, employer-aligned	90-day prototype	High	Cohort or subscription
Cohort program	Community, accountability, high ticket	90-day prototype	Medium	Cohort
Toolkit or templates	Practitioners who implement fast	Quick win	High	One-off or bundle
Playbook or guide	Teams and enterprise buyers	90-day prototype	Medium	One-off or enterprise
On-demand research clinic	Senior buyers with specific problems	Quick win	Medium	Per session or retainer
Decision-support tool	Complex decisions, regulated sectors	Multi-quarter build	High	Subscription or enterprise
Enterprise training package	Corporate L&D buyers	Multi-quarter build	Medium	Enterprise licence
Certification program	Professional development, CPD	Multi-quarter build	Medium	Cohort or annual
Coaching bundle	High-touch work	Quick win	Low	Retainer or cohort

Bundle by outcome, not by topic. A product that promises a measurable deliverable, such as "publish a paper" or "ship a prototype", sells better than a title that describes a subject. Pair a micro-credential with a cohort upsell and you have a credential-first funnel with recurring revenue built in.

Read how to package your IP into digital products before you settle on a format.

---

How do you pick the right product for a £1M+ business?

Score each candidate from 1 to 5 on five criteria, then total it.

1. Founder time saved. Does this run without you once it is built?
2. Repeatability. Can you run it again next quarter with minimal rework?
3. Margin. Does it hold 70%+ gross margin at your target price?
4. Enterprise demand. Could a corporate buyer take ten seats or more?
5. Data for AI. Does it produce signals an agent can act on?

Do this:

• Pick the type that scores highest on founder time saved.
• Prototype at N1 or N2 before you commit to a full N3 build.
• Price with a credential-first funnel. Micro-bundle entry, cohort upsell.

Not this:

• Do not make a coaching bundle your main scale vehicle. It scores low on repeatability and produces almost no usable data.
• Do not launch an N4 program before you have validated demand at N2 or N3.

Worked example. A leadership consultant scores a micro-credential at 22 out of 25 and a coaching bundle at 11. The micro-credential wins because it saves founder time, repeats every cohort, and produces assessment data an agent can use to personalise feedback. The expert-to-AI service playbook runs the full version of that exercise.

---

How do you make knowledge products AI-ready?

Structure is everything. An agent cannot personalise a product that lives in scattered files. Think of a kitchen. Every ingredient needs a labelled container before a chef can cook to order.

Content architecture

Chunk each lesson unit into discrete blocks. One learning objective, one explanation, one worked example, one assessment prompt. Keep each chunk under 400 words. Store them in one structured hub, not across a dozen folders.

Metadata model

Every unit needs these fields so an agent can select and personalise it:

Field	Example value
Unit ID	MC-L2
Learning objective	"Identify three pricing signals in a client brief"
Prerequisite unit	MC-L1
Difficulty level	Intermediate
Outcome tag	pricing-decision
Assessment rubric ID	RUB-04
Canonical prompt	"You are a pricing coach. Given the learner's brief, identify..."
Last reviewed	March 2026

This is the backbone for every agent in your system. Our guide to knowledge architecture for AI covers how to set the schema up properly the first time.

The agents that deliver it

This is where most educators stall. They structure the content, then reach for a generic chatbot to deliver it, and the result sounds like a chatbot rather than like them.

What works is a small set of AI employees, each with one job and each carrying your method's language and standards:

• Authoring agent: drafts new chunks from your canonical prompts and IP notes.
• Delivery agent: selects and sequences chunks based on learner signals.
• Assessment agent: scores responses against your rubric and flags gaps.
• Coaching agent: writes personalised feedback in your method's voice.

We build these with Claude Code, because it lets a non-technical founder ship a working agent in days and keeps the logic in files you own rather than inside someone else's platform. Devwiz covers the retrieval side of this in how to build a RAG pipeline, which is the mechanism that lets an agent answer from your material instead of from general training data.

AI personalisation in education works best when agents get clean telemetry. Completion rates, assessment scores, re-attempt patterns. Log these per unit, not per course.

Pro Tip: Run your first cohort with a delivery agent and an assessment agent only. Add the authoring and coaching agents in cohort two, once the data model has stopped moving.

---

What does a realistic build roadmap look like?

Phase	Timeline	Key activities	Who is needed
Discovery and IP mapping	Weeks 1 to 2	Audit existing content, define N1 objects, set the metadata schema	Founder plus a strategist
Prototype build	Weeks 3 to 6	Build 3 to 5 N2 units, configure delivery and assessment agents, set up telemetry	Founder, developer, instructional designer
Pilot cohort	Weeks 9 to 12	Run the first cohort, collect signals, iterate	Founder plus delivery team
Scale decision	Week 12	Review the numbers, scale or pivot	Founder

What a good prototype looks like:

• Completion rate above 70%
• Net Promoter Score above 40
• Founder hours per learner below two per week
• At least one enterprise enquiry out of the cohort

The case for scalable systems explains why the 90-day window matters at this revenue level.

---

Why fast prototypes beat big launches

Most educators at £1M+ already have the IP. The bottleneck is not ideas. It is the gap between a good method and a system that delivers it without you in the room.

The usual advice is to build the whole program, then launch. That takes six to nine months, costs real money, and gives you exactly one data point. A 90-day prototype gives you learner signals, a working agent stack and a revenue cycle before you commit to the full build.

The products that scale fastest are not the most elaborate. They are the most structured. A well-chunked micro-credential with clean metadata will beat a sprawling course with no agent integration, every time. Structure is what makes your method legible to a machine, and legible to a new team member too.

---

Scale your IP with the 90-day Program

You have a proven method. The question is whether the business can deliver it without you carrying every cohort personally.

The 90-day Program for online educators maps your IP, builds the agent network that carries it, and runs your first automated cohort. More output, fewer founder hours, and a repeatable system your team can run.

It suits you if:

• Your education or consulting business earns £1M+ a year.
• You have a proprietary framework you want to scale.
• You are founder-dependent right now and want that to change.
• You want a working prototype, not a strategy document.

Next step: Take the IP assessment to see which product type fits your method.

---

Useful sources

• PAHO Virtual Campus: types of educational resources, the learning-object classification and the N1 to N4 granularity levels used above.
• Anthropic: Claude Code documentation, the build tool behind the agent roles described here.
• Devwiz: how to build a RAG pipeline, the retrieval mechanism that lets agents answer from your own material.

---