Coalescence versus Obsolescence

The most valuable work of the next decade will be done neither by humans alone nor by AI alone, but by people who know exactly what each brings. AI can now draft, analyse, summarise and calculate at a speed no person can match. Yet it has no stake in the outcome, no body in the world and no life to lose. That gap is not a weakness to be engineered away. It is the foundation of a partnership.

This piece sets out 10 attributes that remain distinctly human, 10 more that AI can help people sharpen, and how the two combine into a genuine force multiplier.

These are not skills AI is merely bad at today. They spring from having a body, a life, relationships and consequences.

1. Lived experience and wisdom from failure — knowledge earned by living through events, and lessons carved by real mistakes with real consequences.

2. Embodiment and physical presence — understanding the world through touch, fatigue and senses; being in the room, shaking a hand, sitting with someone who is ill.

3. Consciousness and genuine empathy — the felt experience of joy, grief and fear, and the ability to feel alongside another person rather than model them.

4. Moral accountability and responsibility to others — being answerable for a decision, and owing a duty to family, colleagues, customers and community.

5. Personal stakes and mortality — having reputation, livelihood and a finite life on the line, which shapes priorities and urgency.

6. Purpose, intent and intrinsic motivation — deciding what something is for, and wanting it for its own sake without being prompted.

7. Trust, love and loyalty — relationships deepened over years, and commitments that hold when they are inconvenient.

8. Intuition and courage — the gut feeling that something is wrong before the data shows it, and acting despite personal risk.

9. Taste — a personal, evolving sense of what is good, beautiful or right.

10. Cultural belonging and humour — membership of a community with its codes and unspoken rules, and the wit and timing rooted in shared life.

These are also human capacities, but here AI acts as a tutor, sparring partner or extra pair of hands. The person stays in charge; the AI widens the range.

1. Curiosity and learning — a patient, personal tutor in any subject, so each question leads to the next instead of stalling.

2. Memory and research — instant recall of notes and past decisions, and sources gathered and summarised in minutes, not days.

3. Analytical and critical thinking — testing assumptions, running the numbers, arguing the opposite case and stress-testing beliefs.

4. Creativity — generating dozens of options so the human can choose and refine.

5. Communication and language — drafting, editing and adjusting tone for any audience, with translation across languages.

6. Decision-making and planning — laying out options, trade-offs and second-order effects, then breaking goals into steps and timelines.

7. Focus and self-reflection — offloading routine admin, and a non-judgemental space to think aloud and examine reasoning.

8. Technical skill and pattern recognition — writing code, building models, using unfamiliar tools and surfacing trends in data too large to read by hand.

9. Leadership and negotiation — preparing difficult conversations and feedback, rehearsing arguments and anticipating the other side.

10. Practical life expertise — understanding health information, modelling personal finances and step-by-step guidance on repairs and making.

A force multiplier makes the same effort produce far more effect. Human plus AI works this way because each side fills the other's gaps.

The human sets direction; AI supplies velocity. Purpose, taste and intent decide what is worth doing. AI compresses the time to do it from weeks to hours.

The human carries judgement; AI carries breadth. AI can scan a thousand documents. Only a person can decide which three matter and stake their name on it.

The human owns accountability; AI absorbs routine. Freed from drafting, formatting and searching, people spend more of their day on decisions only they can make.

The human brings trust; AI brings preparation. A doctor, adviser or salesperson walks into the room better briefed, so the human relationship is spent on what matters.

The human learns; AI accelerates the loop. Faster feedback means more cycles of trying, failing and improving in the same time.

The multiplier only works when each side stays in its lane. Hand AI the judgement and you get confident errors. Refuse AI the routine and you get exhausted experts. The skill of the coming decade is knowing where the line sits.

The effect is strongest wherever expert judgement is scarce and preparation is time-consuming.

Healthcare — What the human brings: Bedside manner, clinical judgement, accountability. What AI adds: Scan analysis, literature review, note-taking. The multiplied result: More time with patients, fewer missed findings.

Education — What the human brings: Mentorship, motivation, belonging. What AI adds: Personal tutoring at each pupil's pace. The multiplied result: Every learner gets individual attention.

Law — What the human brings: Advocacy, ethics, client trust. What AI adds: Document review, precedent search, drafting. The multiplied result: Faster, cheaper access to justice.

Science and research — What the human brings: Hypotheses, intuition, curiosity. What AI adds: Data analysis, simulation, literature synthesis. The multiplied result: Shorter path from question to discovery.

Business and leadership — What the human brings: Vision, risk appetite, relationships. What AI adds: Market analysis, forecasting, reporting. The multiplied result: Better decisions made sooner.

Software and engineering — What the human brings: System design, product taste, ownership. What AI adds: Code generation, testing, debugging. The multiplied result: Small teams building what once took large ones.

Creative arts — What the human brings: Voice, taste, emotional truth. What AI adds: Drafts, variations, production tools. The multiplied result: More ambitious work from a single creator.

Trades and crafts — What the human brings: Hands-on skill, physical judgement. What AI adds: Diagnosis, specifications, instructions. The multiplied result: Better repairs and fewer costly mistakes.

Public service — What the human brings: Democratic legitimacy, empathy. What AI adds: Casework triage, policy modelling. The multiplied result: Faster, fairer services for citizens.

Personal life — What the human brings: Values, relationships, priorities. What AI adds: Planning, finances, health information. The multiplied result: Better-informed choices about one's own life.

Nowhere is the choice between coalescence and obsolescence sharper than inside the enterprise technology estate. Every large organisation runs on systems built over decades, by people who understood not just what the code does but why it does it.

Where the human 10 matter most

Moral accountability and responsibility to others — someone must sign off a production change, a security exception or a data migration, and answer for it.

Trust built over time — CIOs, architects and account managers hold relationships with users, regulators and suppliers that no model can inherit.

Lived experience and wisdom from failure — the engineer who remembers the 2009 outage knows which batch job must never run before the ledger closes.

Intuition and courage — sensing that a vendor's roadmap is hollow, or having the nerve to stop a failing programme.

Taste and purpose — deciding what the product should be, not just how to build it.

Where the AI-assisted 10 transform the work

Technical skill and pattern recognition — AI reads legacy code such as COBOL, PL/I and green-screen workflows, explains it and proposes modern equivalents.

Memory and research — searching years of tickets, change logs and documentation in seconds.

Communication and language — turning terse specifications into clear documentation for new joiners and auditors.

Analytical reasoning and planning — mapping dependencies, sequencing migrations and modelling the risk of each step.

Focus — removing routine testing, reporting and ticket triage so senior people spend their time on design and decisions.

Together, a small team of experienced people working with AI can modernise, document and support estates that once needed hundreds.

The tribal knowledge problem

The most valuable asset in many enterprises is written down nowhere. It lives in the heads of long-serving staff: why a field is padded with zeros, which customer depends on an undocumented screen, why a nightly job is scheduled at 02:17 and not 02:00. This is tribal knowledge.

AI is still at an embryonic stage inside the enterprise. Public models know the internet; they do not know your systems, your exceptions or your history. That knowledge can only come from the people who hold it. If it is not captured now, it will be lost for good.

The risk is acute for three reasons:

The workforce is retiring. Many of the engineers who built core platforms are at or near the end of their careers.

Cost-cutting meets AI hype. Some organisations are releasing experienced staff on the assumption that AI will fill the gap. It cannot fill a gap it was never taught.

AI multiplies what it is given. Feed it accurate institutional knowledge and it scales expertise across the organisation. Feed it nothing and it confidently invents answers.

This is where obsolescence truly lies. Not the human made obsolete by AI, but the enterprise made obsolete by losing the knowledge that made it work.

Coalescence in practice

The answer is to use AI as the capture mechanism while the experts are still here.

Pair veterans with AI. Record walkthroughs and interviews; let AI transcribe, structure and cross-reference them against the code.

Annotate the estate. Have AI draft explanations of legacy programs, then have the people who wrote them correct and approve them.

Record the why, not just the what. Capture decisions, exceptions and past incidents as decision records the AI can retrieve later.

Build a private knowledge base. Ground enterprise AI in this verified material, so its answers reflect your reality, not general guesswork.

Keep humans accountable. Experts validate what is captured; AI makes it searchable, teachable and durable.

Done well, the knowledge of one retiring engineer becomes available to every new joiner, on demand, for years to come. That is coalescence: human experience and machine scale fused into something neither could achieve alone.

Before any knowledge is preserved, one question must be answered honestly: is the business, system or service it supports still viable in the modern economy? That can no longer be assumed.

AI is collapsing the cost of building software, compressing margins and resetting what customers expect. Products that were defensible five years ago can be rebuilt in months. Licence models, support contracts and integration services that once looked permanent are being repriced or bypassed. Longevity is not proof of future value.

Viability should be tested, not presumed. The core questions are:

Market — is there still sustained demand, at a price that supports the business?

Defensibility — what would stop a smaller, AI-enabled competitor rebuilding this quickly?

Strategic fit — is the system core to the organisation's future, or a candidate for retirement or replacement?

Economics — does the cost of running and supporting it justify the value it delivers?

Embedded value — even if the platform has no future, do the business rules and domain knowledge inside it still matter?

The full viability assessment framework is set out in companion documents. Its outcome shapes everything that follows.

Once viability has been confirmed, the path forward becomes clear, and the knowledge effort can be scaled to match:

Viable — preserve: The full body of tribal knowledge: system, process, customer and domain. Approach: Apply GRAFT in full and invest in AI-grounded knowledge as a lasting asset.

Transitional — preserve: Business rules, domain logic and customer context, not platform minutiae. Approach: Apply GRAFT selectively to carry the knowledge into the replacement.

Not viable — preserve: Only what is needed for an orderly exit: data, compliance and obligations. Approach: Capture the minimum, then redirect people and investment to what has a future.

This is the discipline that separates coalescence from sentiment. Preserving knowledge for a business with no future is simply a slower form of obsolescence. Preserving it for one that is viable is among the highest-return investments an enterprise can make.

Grafting is the right metaphor. A gardener takes a living cutting, the scion, and joins it to a healthy rootstock so it takes root and grows. Tribal knowledge is the scion: living, specific and irreplaceable. The enterprise and its AI are the rootstock. Once viability is confirmed, GRAFT is a ten-step method to make the join take: Gather, Record, Adapt, Feedback, Track.

G — Gather, step 1. Map the knowledge at risk: Identify critical systems, single points of failure and the people who hold what is undocumented, starting with those nearest retirement or exit. Output: Knowledge risk register.

G — Gather, step 2. Elicit from the source: Run structured interviews, screen-share walkthroughs and shadowing sessions. Ask for the why, the exceptions and the past incidents, not just the how. Output: Recorded sessions.

R — Record, step 3. Capture and structure: AI transcribes and organises sessions into runbooks, decision records, FAQs and annotated code. Output: Draft knowledge assets.

R — Record, step 4. Anchor to the estate: Link every item to the system, program, job or process it describes, held in a private, searchable knowledge base. Output: Connected knowledge base.

A — Adapt, step 5. Translate for new hands: Rewrite in plain language and modern terms, with versions for engineers, support, operations and business owners. Output: Role-based guides.

A — Adapt, step 6. Ground the AI: Load the verified knowledge into enterprise AI so it answers from your reality, not general guesswork. Output: Knowledge-grounded AI assistant.

F — Feedback, step 7. Validate with the expert: The original knowledge holder reviews and corrects AI drafts before anything is published. Human accountability seals the graft. Output: Expert-approved content.

F — Feedback, step 8. Test in real work: New joiners and support teams use it on live tasks, flagging gaps, errors and missing context. Output: Gap and correction log.

T — Track, step 9. Measure the take: Track coverage of critical systems, usage, answer accuracy, time-to-competence for new staff and incidents avoided. Output: Knowledge health dashboard.

T — Track, step 10. Keep it alive: Assign owners, set review cycles and update content whenever systems change. Living knowledge needs tending. Output: Sustained knowledge asset.

The sequence matters. Gather and Record must happen while the experts are still in the building. Adapt and Feedback turn raw recollection into trusted, usable knowledge. Track ensures the graft takes and keeps growing. Skip the human validation in step 7 and the graft fails: AI will spread errors as confidently as truths.

The question is not whether AI will replace people, but which people and organisations will learn to work with it. The first 10 attributes are why humans remain essential. The second 10 are where AI makes them stronger. Coalesce the two, and an individual can do the work of a team, and a team the work of an organisation. Fail to, and above all fail to capture the knowledge held by today's experts while they are still here, and obsolescence follows, not for the people, but for the enterprise itself.