Free worksheet

Place every AI idea by how long it takes to work

The Opportunity Pyramid sorts AI ideas by one thing: time from decision to deployment. Put your ideas on it, count the tiers, and the shape tells you where to start.

Download the worksheetHow to use it
The AI Opportunity Pyramid

Four tiers, measured in time

Not a ranking of value or sophistication. A ranking of distance from deployment. Choose a tier to see what sits in it.

The AI Opportunity PyramidFour tiers by time to deploy

Choose a tier to explore it

6–18 monthsNew capability

Embedded and bespoke AI

AI inside the product you sell, custom models for your own process, small models on your own devices, and agents that run a function within set limits. Research: you are finding out whether it can be done, not buying it.

MonthsSkills

Integrated enterprise AI

Systems that cut across functions and depend on data being joined up: a data project with AI on the end of it. The AI is often the smaller part; most of the work is getting the data into one place, in one shape.

WeeksQuality

Specialist functional AI

Tools built for one function with AI inside them, configured and connected to one or two systems. The decision is which product, not whether to build. The work is done better, more consistently, with fewer errors.

DaysEfficiency

Co-pilots and AI assistants

Off-the-shelf tools on a licence: the office suite’s assistant, a general model, a meeting transcriber, a coding assistant. No data to prepare and nothing to connect: the work is governance and training. The same work, done faster.

Example from

Industrial manufacturing · Embedded AISelf-diagnosing equipment, with models running on the products you make.e.g. built on Azure IoT Edge or AWS IoT Greengrass

Publishing and media · Embedded AIA recommendation model of your own, across every title you publish.e.g. built on Google Cloud, AWS or Azure machine learning

Ecommerce and digital agencies · Embedded AIA support agent that answers from each client’s own code, and hands over to a person when unsure.e.g. built with Claude, on your helpdesk and repositories

Retail and consumer brands · Embedded AIA recommendation model trained and owned in-house, on your own catalogue and customers.e.g. built on Google Cloud, AWS or Azure machine learning

Food, drink and FMCG · Embedded AIA demand model of your own, trained on your history and built into weekly planning.e.g. built on Azure, AWS or Google Cloud machine learning

Healthcare, pharma and animal health · Embedded AIAI inside a diagnostic or monitoring product, developed and validated under device regulation.e.g. built on Azure or AWS machine learning, with your quality system

Financial and business services · Embedded AIAgents that prepare routine filings within set limits, for a professional to sign off.e.g. built with Claude or OpenAI models, on Azure or AWS

Charities and membership bodies · Embedded AIYour own guidance service for the people you support, built and tested with safeguards.e.g. built with Claude or OpenAI models, with your own testing

Sport, leisure and tourism · Embedded AIRecommendations of your own in the club or venue app, from content to seat upgrades.e.g. built on Google Cloud, AWS or Azure machine learning

Higher education · Embedded AIA course tutor of your own inside the learning environment, tested with academics.e.g. built with Claude or OpenAI models, in Moodle or Canvas

Industrial manufacturing · Integrated enterprise AIPredictive maintenance and quality root cause, from line, test and ERP data joined up.e.g. SAP or Microsoft Dynamics data, joined in Microsoft Fabric or Databricks

Publishing and media · Integrated enterprise AIPredicting which subscriber will lapse, and which offer will keep them.e.g. your subscription data, joined with analytics in Google BigQuery

Ecommerce and digital agencies · Integrated enterprise AIAnswers for the support team from every client’s code, documents and past tickets.e.g. Claude or OpenAI models, with search across your repositories and helpdesk

Retail and consumer brands · Integrated enterprise AIDemand forecasts on stock, promotions and sales, joined up across your channels.e.g. a forecasting tool on Microsoft Fabric or Google BigQuery, with your ERP

Food, drink and FMCG · Integrated enterprise AIPromotion forecasts from orders, trade promotions and retailer sales data joined up.e.g. a forecasting tool on Microsoft Fabric or Snowflake, with your ERP

Healthcare, pharma and animal health · Integrated enterprise AIA compliance assistant that answers from SOPs, product files and guidance, with its sources.e.g. Azure AI Search or Amazon Bedrock, over your document systems

Financial and business services · Integrated enterprise AIAnswers grounded in client files and professional guidance, with access that follows the client.e.g. Azure AI Search with Claude or OpenAI models, over your document systems

Charities and membership bodies · Integrated enterprise AIMember and donor data joined up, so lapses are spotted and the next ask is right.e.g. Salesforce or Microsoft Dynamics, with your fundraising and events data

Sport, leisure and tourism · Integrated enterprise AIOne view of each fan across tickets, retail and membership.e.g. Salesforce or Microsoft Dynamics, joined with ticketing and retail

Higher education · Integrated enterprise AIA student-service assistant that answers from policies and the student record, for each student.e.g. Azure AI Search with Claude or OpenAI models, joined to the student record

Industrial manufacturing · Specialist functional AISupplier invoices and delivery notes read and matched automatically.e.g. the invoice capture in Microsoft Dynamics 365, SAP or Sage

Publishing and media · Specialist functional AIArticles tagged automatically, and AI search across the archive for the newsroom.e.g. Azure AI Language or Google Cloud Natural Language, with your CMS

Ecommerce and digital agencies · Specialist functional AISupport tickets tagged, prioritised and routed in the helpdesk you already use.e.g. the AI in Zendesk, Freshdesk or Jira Service Management

Retail and consumer brands · Specialist functional AIAI search on the store that understands what shoppers mean.e.g. an AI search product on Shopify Plus, Adobe Commerce or BigCommerce

Food, drink and FMCG · Specialist functional AIConsumer complaints read, tagged and routed to the quality team.e.g. the AI in Zendesk, Freshdesk or Salesforce Service Cloud

Healthcare, pharma and animal health · Specialist functional AIRegulatory and quality documents searched and compared by the regulatory team.e.g. the AI in Veeva Vault or MasterControl

Financial and business services · Specialist functional AIContract review, or bookkeeping capture, for one practice area.e.g. Luminance or Harvey for legal; Dext or AutoEntry for accountancy

Charities and membership bodies · Specialist functional AISupporter enquiries sorted and routed to the right team.e.g. the AI in Zendesk, Freshdesk or Salesforce Service Cloud

Sport, leisure and tourism · Specialist functional AIFan and visitor enquiries answered and routed in the helpdesk.e.g. the AI in Zendesk, Freshdesk or Salesforce Service Cloud

Higher education · Specialist functional AIStudent enquiries answered and routed in the student-services helpdesk.e.g. the AI in TOPdesk, Zendesk or Salesforce Service Cloud

Industrial manufacturing · Co-pilots and AI assistantsSupplier emails drafted and technical datasheets summarised.e.g. Microsoft 365 Copilot, ChatGPT or Claude

Publishing and media · Co-pilots and AI assistantsHeadlines, summaries and social posts drafted from your articles.e.g. Microsoft 365 Copilot, ChatGPT or Claude

Ecommerce and digital agencies · Co-pilots and AI assistantsBriefs, client updates and first code reviews drafted.e.g. GitHub Copilot, Claude or ChatGPT

Retail and consumer brands · Co-pilots and AI assistantsProduct descriptions and campaign copy drafted in your brand voice.e.g. Shopify Magic, Microsoft 365 Copilot or Claude

Food, drink and FMCG · Co-pilots and AI assistantsTrade presentations, specifications and supplier emails drafted.e.g. Microsoft 365 Copilot, ChatGPT or Claude

Healthcare, pharma and animal health · Co-pilots and AI assistantsInternal reports, literature summaries and training material drafted, with no patient data.e.g. Microsoft 365 Copilot or Claude for Enterprise

Financial and business services · Co-pilots and AI assistantsMeetings summarised, client letters drafted and first-pass research done.e.g. Microsoft 365 Copilot, Copilot in Teams or Claude

Charities and membership bodies · Co-pilots and AI assistantsAppeals, grant bids and member newsletters drafted.e.g. Microsoft 365 Copilot, ChatGPT or Claude

Sport, leisure and tourism · Co-pilots and AI assistantsMatch reports, fan emails and social posts drafted in minutes.e.g. Microsoft 365 Copilot, ChatGPT or Claude

Higher education · Co-pilots and AI assistantsCommittee papers, course materials and emails drafted and summarised by staff.e.g. Microsoft 365 Copilot Chat, Gemini for Education or Claude

How to use it

Four steps, one afternoon

  1. List every AI idea you have heard in the last year, one per line. Include the dull ones; they are often the first wins.
  2. Place each by time to deploy, with no optimism: days, weeks, months, or six to eighteen months. If you do not know, it is months.
  3. Add a data position to each: low dependency, in hand, siloed, partial or absent. Ask the person who would have to fetch the data. See the readiness checklist for what each means.
  4. Count the tiers. The worksheet does this for you, and draws the shape.
Reading the shape

What the counts tell you

  • A thin base

    Few ideas in the days tier usually means the tools are blocked. Opening governed access is the first move.

  • A heavy middle

    If more than half the ideas need months, you have a data project before you have an AI project. Fund the data once, for all of them.

  • A small apex

    The six-to-eighteen-month ideas are few and often matter most. Fund them as research, not as a business case.

  • Sequence by readiness

    Start with what can be done now, not with what is most exciting. Nothing is dropped; the order is set by what is ready.

Questions

Using the worksheet

Isn't the top of the pyramid the most valuable?

Not necessarily. An assistant used by four hundred people can be worth more than a bespoke model. The tiers measure time to deploy, not value.

Can an idea move between tiers?

Yes. What had to be built a few years ago can often be bought and configured today. Revisit the placement every six months.

What comes after the worksheet?

Score the ideas you want to pursue with the use-case scorecard, then write a pilot brief for the first one.

See the whole picture

CONVERGE chapter 4 explains the Opportunity Pyramid in full, with a real programme's distribution across the tiers. Or we can place your ideas with you, in a discovery workshop.

CONVERGE, the bookBack to the hub