The upskilling engine
Deep Academy.
Organisations are buying AI while their people stay in a pre-AI operating model. The gap that produces is not a technology gap. It is a capability gap, and it is the one nobody budgets for.
With Crossroads Catalyst · VITAL Expert · Ludic / SmartLab
01The problem
Generic AI training teaches people about AI.
It does not teach a domain expert how to orchestrate AI inside their own workflow, their own regulatory environment, their own professional standard. A pharmaceutical marketer, a regulatory writer and a medical science liaison need fundamentally different competencies — and most programmes treat them identically.
So the tools land, the training happens once, and six months later the team works the way it always did. That is not a failure of attention. It is a failure of design: the training was never about their work.
Deep Academy exists to close that gap, and it is the reason the other three engines produce anything durable. A platform nobody adopts is a licence fee.
02The distinction
From literacy to fluency.
AI literacy asks: do you know what AI can do?
AI fluency asks: can you direct several AI systems to complete complex, domain-specific work at a standard you would put your name to?
The second question is the one that matters, and almost nothing on the market answers it. The future of this work is not people using tools. It is domain experts orchestrating systems — and then exercising the judgement that decides whether the output is good enough to sign.
The word we use is orchestration. The skill underneath it is judgement.
Deep Academy turns domain experts into Deep Workers: professionals who combine years of specialised knowledge with the ability to direct AI systems, evaluate what comes back with expert scepticism, and take responsibility for the result.
03The word
Four meanings of “deep”.
Deep expertise
Domain mastery built over years of practice. We are not replacing it — the whole method depends on it.
Deep work
Focused, undistracted effort. Orchestration is supposed to buy more of it, not fragment what is left.
Deep learning
A shift in mental model, not a skills checklist. People have to change how they think about their own work.
Deep integration
Capability embedded in real work, not demonstrated in a training environment and forgotten.
04The Deep Worker
Where breadth meets depth.
A T-shaped profile. The vertical bar is domain expertise — years of specialised knowledge in one function. The horizontal bar is orchestration breadth: context engineering, multi-agent workflows, quality assurance, governance, tool integration.
Neither alone is worth much. A domain expert without orchestration is capacity-bound. An AI specialist without domain context produces confident work that a professional would not sign.
High domain · low orchestration
Traditional expert
Deep domain knowledge, little leverage. Capacity is a function of hours available, and the method leaves when they do.
★ The Deep Worker
Deep domain + orchestration
Expertise that scales past the individual's calendar — and a method that persists after the person moves on.
Low domain · low orchestration
Generalist
Neither depth nor leverage. Shallow impact in a field where the details decide the answer.
Low domain · high orchestration
AI specialist
Orchestration without domain context. Fast, fluent, and unable to tell when the output is subtly wrong.
Deep Academy moves people from the top-left to the top-right — preserving the expertise, adding the leverage. It does not try to create expertise, which is why it works on senior people and not on graduates.
05Tracks
Nine tracks, one per function.
Each takes a traditional role and builds its orchestrating equivalent. The domain content differs completely between tracks — which is the point, and the reason a single generic curriculum cannot do this.
Deep MedicalMedical affairs & MSL
Orchestrating literature synthesis and scientific engagement across a therapy area, with the evidence graded and the reasoning kept.
Deep MarketeerBrand & commercial
Generating and reviewing content variants at pace, with the claim each message can support written down beside it.
Deep Market AccessHEOR & value
Accelerating evidence synthesis and value-story construction, with every source traced to its origin and date.
Deep RegulatoryIntelligence & review
Coordinating landscape and precedent analysis, and reviewing evidence packs. Not the writing of filed content — see the boundary below.
Deep R&DDiscovery & development
Structuring target and protocol analysis so options are compared on the same basis rather than argued from memory.
Deep StrategyPortfolio & business development
Orchestrating multi-source analysis and scenario construction, with observable triggers rather than probabilities.
Deep AnalystData science & BI
Building intelligence pipelines that stay current, and knowing which signals are worth a person's attention.
Deep HRTalent & people operations
Designing skills-based workforce architecture for an organisation whose capability mix is changing.
Deep FinanceFP&A & forecasting
Structuring forecasting and reporting workflows so the assumptions are visible and re-runnable.
06Programme journey
Five phases, and each one ends in something you can show.
The journey is designed with you before it starts. Phase 0 sets the baseline and selects the track; everything after it is built on what that diagnostic finds.
Diagnostic & designEstablish the baseline
Track selection and a customised learning plan carrying the baseline scores, with access to curated content on Smart Lab.
- AI adoption self-assessment — mindset, skillset, toolset, modeset
- Role mapping and track selection
- Use-case selection and success criteria
- Cohort design — cadence, coaching model, artefact expectations
FoundationThe shared core
A common language, the quality rules, a first validated micro use case that applies to daily work, and a prompt library.
Track specialisationCapability by function
Track workflows, templates and artefacts for two use cases.
Applied practicumThe methods on real work
A validated capstone and a measurable business application, produced with coaching and review.
Mastery & scaleOrchestration and enablement
An operating playbook, a multi-team workflow, and an internal champion layer.
07Curriculum design
Built from a persona ontology, not from a tool list.
The curriculum is designed from a systematic mapping of pharmaceutical roles and how each role’s work profile shifts when intelligent systems take on different task categories. We mapped forty-three distinct personas across medical affairs, commercial, regulatory, R&D, market access, HR and finance.
The pattern was consistent enough to build on: the roles that hold up are the ones that move from executing to orchestrating, and the skill that decides the outcome is evaluation rather than generation.
That mapping drives track design, use-case selection, competency benchmarks and certification levels. We do not teach AI in the abstract — we teach it through the lens of how one specific role’s work is changing.
08Delivery
Three partners, one programme.
Strategy, transformation and domain expertise. Designs the curriculum, writes the domain content, and makes sure every track reflects how the work is actually done.
The environment where participants build and test, running on the same substrate their production work would use — so what they learn transfers.
Learning delivery, change management and scale. Running cohorts across a global organisation is a different discipline from designing one.
Architects, sandbox, stage. Curriculum without an environment is a slide deck; an environment without change management is a licence nobody uses.
09Certification
Three levels, earned on real work.
Assessed on a capstone built from the participant’s own deliverables, reviewed by a practitioner in their function. Not a completion certificate — attendance is not the thing being certified.
Deep Foundation
Understands what these systems can and cannot do, and works competently inside a framework someone else designed.
Deep Practitioner
Designs and runs domain-specific workflows independently, and knows when to stop and involve a person.
Deep Master
Orchestrates coordinated systems across functions, and can teach the discipline to others.
One role sits above these and is defined separately: the Domain Orchestrator — the named person who owns a codified service and is accountable for its quality. Every organisation running this at scale needs one, and most discover that eighteen months late.
10Start
Three steps, and the first is short.
DiscoveryWhere the gap actually is
Map which roles need which capability, and in what order. Usually reveals that the constraint sits somewhere other than where the budget was pointed.
Track prioritisationWhere to start
Identify the one or two tracks with the highest chance of visible change, because the first cohort decides whether there is a second.
Pilot designWith the success measure agreed first
Configure the first cohort, with what counts as success written down before it starts rather than reconstructed afterwards.
Deep Academy runs as enterprise cohorts, with tracks configured to your priorities and your regulatory environment.