Planning to Age Well, with AI

Mind, muscle and money — the OWL framework, and a first plan you can start this week

Author

Behzad Samadi

Published

September 8, 2026

Four people of different ages climbing a golden staircase, with the headline: for aging well, you need three M's — Mind, Muscle and Money

Aging well is not one problem. It is three, and they compound the same way.

You need mind — enough cognitive reserve that you are still the one making the decisions about the other two. You need muscle — enough strength, balance and aerobic capacity that you can carry your own groceries up your own stairs at eighty. And you need money — enough capital that your later years are not governed by the price of things.

The three Ms are the core axiom of the OWL framework — Optimal Wealth & Longevity — and it exists because of a specific failure in how retirement is normally planned: cognitive health, physical health and financial engineering are handled by different people, in different rooms, on different timescales — a family doctor with eleven minutes, a physiotherapist, and a planner. Optimise only the mind and you have nothing to spend the clarity on, or the strength to use it. Optimise only the body and you arrive at eighty strong, anxious, and short of options. Optimise only the money and you can fund a very comfortable confinement.

NoteThis is a method, not advice

Nothing here is personalised cognitive, medical or financial advice, and it cannot be — the right answer depends on your income, your health, your obligations and your jurisdiction. Treat it as a way to arrive at good questions and a first draft. Treat any real change in memory, word-finding or planning ability as something to have assessed, never as something to train around. Take the rest of the draft to your doctor — especially before starting resistance training if you have a heart condition, uncontrolled hypertension, or a recent injury — and to a fee-only financial planner.

Symmetrical compounding

The reason the three belong in one plan is that they obey the same arithmetic. Mind, muscle and money all compound, all decay when unattended, and all fail in a way that is sudden rather than gradual.

Mind Muscle Money
Compounding Education, novel skills and social ties compound into cognitive reserve — capacity that keeps functioning even as underlying pathology accumulates. Early strength training compounds bone density, joint integrity and muscle reserve. Early, regular contributions compound over multi-decade horizons.
Decay Unused capacity narrows the same way — processing speed and working memory drift downward from midlife, faster once sleep, isolation or vascular risk factors go unmanaged. Biological inflation: untrained adults lose muscle mass every year past 40. Inflation erodes unhedged purchasing power by roughly 2–4% a year.
Failure mode Mild cognitive impairment progressing toward dementia — the failure mode that can take away your authority over the other two. Sarcopenia, frailty, loss of independence, a fall that breaks something. Portfolio depletion during a high-inflation or high-volatility drawdown.
Protection Aerobic exercise, sleep, social contact and novel learning, plus the vascular risk factors — hypertension, hearing, diabetes — that double as heart and brain protection. Progressive overload, power training, and enough protein to use it. Guardrail withdrawal rules and defensive allocation.

Biological inflation is the idea worth taking away, and it turns out to apply twice. Everyone accepts that cash left under a mattress quietly loses value. Far fewer people plan for the fact that an untrained body does exactly the same thing, on a comparable schedule — and fewer still notice that an untrained mind does too.

Cognitive reserve behaves less like a stockpile than like a credit line: two people with identical scan-confirmed pathology can present a decade apart in symptoms, and the gap tracks with education, occupational complexity, and how much genuinely novel demand a person keeps placing on their own thinking. Reserve is not acquired once and kept. It is built the way the other two are — on a schedule that starts paying off decades before you need it, and depreciates on the same schedule when neglected.

The Lancet Commission on dementia prevention puts a number on the neglect: an estimated 45% of dementia cases trace to fourteen modifiable factors spread across a lifetime — less education early on; then hearing loss, hypertension, obesity, heavy alcohol use and traumatic brain injury in midlife; then smoking, depression, physical inactivity, diabetes, low social contact, air pollution, untreated vision loss and high LDL cholesterol later. Roughly half that list is not a brain problem at all — it is a heart, an ear and an eye problem, which is a large part of why the muscle and money plans below are already doing some of this work.

The body shows the same arithmetic in numbers most people already sense: commonly cited figures put muscular decline at roughly 3–8% of muscle mass per decade after thirty, accelerating after sixty to something closer to 1–2% per year. Strength falls faster than mass, and power — force produced quickly, the thing that turns a stumble into a recovery rather than a fracture — falls faster still.

The three-legged stool. Traditional financial planning solves only for money. But wealth in a wheelchair is a different asset from wealth on a mountain trail, and wealth you can no longer manage yourself is a different asset again. A plan that skips the body leaves you half prepared; a plan that skips the mind leaves you unable to tell.

The three phases

Retirement is usually modelled as a flat thirty-year line. It is not one. OWL splits it into three phases, each with a cognitive, physical and financial strategy that have to be matched to one another — because cognitive reserve, physical capacity and spending are all closest to their peak at the same time, and only briefly.

Phase 1 — Go-Go (roughly 60–72)

Highest vitality, and the window where discretionary spending buys the most.

  • Cognitively: build reserve deliberately, in genuinely new domains rather than more of what you already do well — a language, an instrument, a professional or creative skill you have never had. This is also the phase to get hearing, vision, blood pressure and blood sugar checked and controlled, while doing so is routine maintenance rather than a crisis, and to set up the legal contingency — power of attorney, a health directive — while capacity is not in question.
  • Physically: build the largest possible buffer. VO₂ max, joint mobility, compound resistance training. Everything later draws down from what is banked here.
  • Financially: peak discretionary expenditure — travel, ventures, active living — deliberately matched to peak physical autonomy. Spending later on what you can only do now is a poor trade.

Phase 2 — Slow-Go (roughly 73–82)

The transition, and the phase where independence is actually won or lost.

  • Cognitively: protect against isolation as deliberately as against a fall. A standing social commitment — teaching, volunteering, a class, a group — is the single intervention with the largest evidence base at this stage. Sleep quality and mood are worth tracking here as first-order signals, not background noise.
  • Physically: balance, proprioception, fall prevention, rotational power, and sit-to-stand speed. The emphasis moves from building to protecting.
  • Financially: reallocation away from high-intensity travel toward home ergonomics, local experiential leisure, and tax-efficient estate structuring.

Phase 3 — Care Support (83+)

Preservation.

  • Cognitively: preserve orientation, communication and remaining capacity through routine, familiar surroundings, and family or caregiver engagement. This is also the phase that proves whether the Phase 1 legal contingency was actually set up — it is nearly impossible to arrange cleanly once it is needed rather than before.
  • Physically: functional movement, bed-to-chair independence, grip strength.
  • Financially: drawdown of ring-fenced care funds and longevity annuities, structured so health expenses do not force the liquidation of core assets.

The value of the phasing is that it makes the coupling explicit. A cognitive marker moving early — a change in planning ability or word-finding — is not only a symptom. Neither is a physical one — grip strength, or sit-to-stand time. Either is a signal to bring forward the legal paperwork, home retrofitting and care provisioning, years before any of it becomes urgent and expensive.

What AI is good at here — and what it is not

Being precise about this matters, because the failure mode is trusting it for exactly the wrong things.

It is good at:

  • Turning intent into structure. “I want to retire comfortably” is not a plan. AI is very good at converting that into named accounts, contribution amounts and review dates.
  • Arithmetic and scenarios. Compounding, drawdown, inflation adjustment, “what if I retire three years earlier”. Check the numbers; the reasoning is usually sound and the arithmetic is not always.
  • Designing novel-skill curricula. A model can keep generating unfamiliar material — a new grammar, a new instrument’s repertoire, a new domain’s fundamentals — long after your own supply of ideas for what to learn next has run out, which matters because novelty, not repetition, is the active ingredient in building reserve.
  • Programme design and progression. Building a twelve-week block, substituting a movement around a bad shoulder, deciding when to add weight.
  • Generating the questions to ask a professional. Arguably its highest-value use. Walk into an advisor’s office with twenty specific questions and you get a different meeting.
  • Adherence. A weekly ten-minute review with a model that has your plan in context is a low-friction way to not quietly abandon it.

It is not good at:

  • Knowing your actual numbers. It will invent plausible ones if you let it.
  • Diagnosing anything cognitive. A conversational model can feel reassuringly sharp with you specifically, because it adapts to your own phrasing. It is not a validated screening instrument, and it will not notice a change in you the way someone who has known you for twenty years will.
  • Medical judgement. It cannot see your bloodwork, your scan or your knee.
  • Regulatory and tax specifics, which change yearly and vary by country. It will state last year’s contribution limit with total confidence.
  • Caring whether you do it. Motivation is still yours.

Where this is heading

The OWL framework describes something more ambitious than a chat window: four cooperating agents. A mindspan agent tracking cognitive markers — processing speed, working memory, sleep quality, social-contact frequency. A healthspan agent ingesting biometrics — wearables, grip strength, sit-to-stand markers, VO₂ trajectories — to model biological rather than chronological age. A wealthspan agent running tax-aware decumulation and Monte Carlo stress tests against horizons past ninety-five. And a cross-orchestrator that harmonises the three: when cognitive markers drift, capital moves toward the legal and financial safeguards — power of attorney, a trusted contact on file — that are cheap to arrange early and nearly impossible to arrange cleanly after the fact; when the biometrics show early mobility decline, capital moves pre-emptively toward physiotherapy and home modification.

None of that is available off the shelf today, and the parts that touch cognition, health and money are exactly the parts that need regulation and auditability before they should be trusted. But it clarifies what to do now: instrument all three sides, keep all three plans in one place, and review them together. The orchestration can start as a calendar reminder and a text file.

Plan one: the mind

The goal is not trivia. It is staying the author of your own decisions — reading a contract, remembering the plan you made, and noticing when something is wrong before someone else has to notice it for you.

There is no WHO minutes-per-week target for this the way there is for exercise, because the strongest interventions are not tasks so much as decisions. The WHO’s own guideline on cognitive decline and dementia grades its evidence, and much of what it can recommend with real confidence is already in the muscle and funds plans below — physical activity, blood pressure control, not smoking, limited alcohol — plus a few more worth listing on their own:

Novel learning Not more of what you are already good at. A new language, instrument, or genuinely unfamiliar skill — novelty is the active ingredient; a harder crossword mostly trains you to be better at crosswords.
Social contact Structured and regular, not incidental. A standing weekly commitment — teaching, volunteering, a class, a group — outperforms passive contact with the same people.
Sleep 7–9 hours. The brain’s waste-clearance system runs mainly during deep sleep; chronic short sleep is one of the more consistently replicated risk markers in the literature.
Vascular risk factors Hypertension, diabetes, hearing loss and vision loss are cardiovascular and sensory problems with an outsized cognitive cost — the largest single block of modifiable risk in the Lancet Commission’s accounting, and mostly a matter of showing up to the appointments.
Aerobic exercise Already in the muscle plan. One randomised trial found a year of aerobic training grew hippocampal volume by about 2% in older adults — a rare case of reversing an age-related loss rather than merely slowing it.

Two things the framework adds that most cognitive-health advice leaves out:

  • The point of the exercise is not the exercise. A puzzle app that gets harder every week is optimising the wrong metric — proficiency at that specific task, not transferable reserve. The evidence favours breadth (a new domain) over depth (a harder version of an old one).
  • This is the plan that protects the other two. A financial plan and a training programme both assume someone is still capable of running them. Set up the legal contingency — power of attorney, a health directive, a trusted contact on file with your bank — while capacity is not in question, in Phase 1, not when it first becomes one. It costs an afternoon now and is nearly impossible to arrange cleanly later.

There is no cheap equivalent of a grip-strength dynamometer for this, and that is worth being honest about: a validated screening instrument (the MoCA is the common one) needs a clinician to administer and interpret it, and a phone quiz that claims to diagnose anything should be treated as entertainment, not a baseline. What you can reasonably track yourself: how many genuinely new skills you picked up in the last year, and how many recurring social commitments are on your calendar. Both are zero for a surprising number of people, and both are fixable this week.

A prompt that gets a useful answer:

I want to build cognitive reserve, not just pass the time. I am 43, work a desk job, and have about three hours a week to spend on something genuinely new. Suggest three unfamiliar skills or domains I could commit to for the next six months — at least one should have a social or structured-group version available near me — and explain, for each, why it counts as novel learning rather than practice at something I already do.

Take any real change in memory, word-finding or planning ability to a doctor before training around it — that is the one place in this whole framework where the right move is a phone call, not a prompt.

Plan two: the muscles

The goal is not aesthetics. It is staying independent — carrying things, climbing stairs, and getting up off the floor unaided at eighty. That takes about three hours a week.

The WHO guidelines for adults 65+ are a reasonable skeleton: 150–300 minutes of moderate aerobic activity a week, muscle-strengthening on 2+ days, and balance training on 3+ days.

A first twelve weeks, deliberately minimal so it survives a real schedule:

Frequency Two full-body sessions a week, 45 minutes. A third is a bonus.
Movements One squat, one hinge, one push, one pull, one carry.
Progression Add weight or a repetition whenever you complete all sets with good form. Progressive overload is the entire mechanism.
Power Once a week, something fast and light. Power declines fastest and is trained least — and it is what prevents falls.
Balance Two minutes daily. Single-leg stand while the kettle boils.
Aerobic Most days, a brisk 30-minute walk. Add one harder session once the habit holds.
Protein Older adults are commonly advised well above the 0.8 g/kg/day adult RDA — often cited in the 1.2–1.6 g/kg/day range, spread across meals. Discuss with your doctor if you have kidney disease.

Two things the framework adds that most training advice leaves out:

  • Muscle is a metabolic organ, not just a mechanical one. Skeletal muscle is the body’s largest glucose sink; maintaining it buffers against metabolic syndrome, which is itself entangled with cognitive decline. Training is not only about not falling.
  • Strength has a balance-sheet consequence. The framework’s blunt version: an hour of resistance training at 55 buys down long-term care liability at 80. Treat the specific dollar figures as directional rather than actuarial — but the direction is not controversial, and assisted living is expensive.

Track two markers, because both predict later independence and neither needs a lab: grip strength (a cheap dynamometer) and sit-to-stand speed (five reps, timed). Measure now so you have a baseline rather than a memory.

A prompt that gets a useful answer:

Build me a 12-week, twice-weekly full-body strength programme. I am 43, have not trained seriously in six years, have a gym with barbells and dumbbells, and an old left shoulder injury that dislikes overhead pressing. Use one squat, one hinge, one push, one pull and one carry per session. Specify sets, reps and weekly progression, and give me the two form cues that matter most for each movement.

Take the output to a coach for one session before loading it heavily.

Giving an AI your numbers

Plan in ratios, not identifiers. Never paste your Social Insurance or Social Security number, account or card numbers, banking credentials, or full medical records — and be aware that anything you paste into a hosted model has left your machine.

The good news is that you do not need to share any of it. Almost every question worth asking is about proportions, so translate before you paste.

Translate to ratios first

Express everything as a multiple of your annual spending, $X. That single absolute number is the one worth giving: it is not an identifier, and every other figure becomes meaningful against it.

  1. Income and assets become multiples of $X. Instead of “I earn $120,000, have $350,000 in my RRSP, and spend $60,000 a year” → “My annual spending is $X = $60,000. Salary 2X. Registered retirement portfolio 5.8X.”
  2. Accounts become functional codes. “Scotiabank TFSA #10293” → TFSA_Equities. “Sun Life Group RSP” → RSP_Employer_Match.
  3. Debts become a rate and a size, highest rate first. “$4,500 on a Visa at 19.9% and $412,000 on the mortgage at 4.8%” → “Debt A: 0.07X at 19.9%, revolving. Debt B: 6.8X at 4.8%, mortgage, 18 years remaining.”

Ordering debts by rate rather than by size is not only anonymisation — it puts them in the order you should actually clear them.

A prompt built from the translation

The profile below is the same person as the example above, translated:

I want to draft a five-year contribution and drawdown sequence.

Ratio-based profile:
- Base annual spending: $X = $60,000
- Debt A (revolving, high-interest): 0.07X at 19.9%
- Debt B (mortgage): 6.8X at 4.8%, 18 years remaining
- [Emergency_Cash]: 0.5X, high-interest savings
- [TFSA_Core]: 1.2X, global equities
- [RRSP_Core]: 2.5X, broad index funds
- Savings capacity: about 18% of gross income

Jurisdiction: Canada. TFSA contribution room is caught up; RRSP unused
deduction limit is roughly 0.25X.

Using the OWL framework, draft a step-by-step contribution
sequence for the next five years. State your assumptions, show the
arithmetic, and flag anything you are unsure of rather than inventing it.
Output as markdown suitable for a file named funds-plan.md.

Then check every contribution limit against the tax authority’s own site before acting on any of it.

Plan three: the funds

A first draft in five moves. The order matters more than the precision.

  1. Measure what a year costs you. Twelve months of actual spending, not an estimate. Everything downstream is a multiple of this number. Most people are wrong by 20–30% on their first guess. A few months of statements and a small local tool will get you a measured figure rather than a guessed one.
  2. Set the floor. Three to six months of that spending, in cash, in a separate account. This is your sequence protection — the buffer that stops an early market downturn forcing you to sell into it. Sequence-of-returns risk is the single most underrated danger in the first years of drawdown.
  3. Clear high-interest debt. Anything above roughly 8–10% is a guaranteed negative return no portfolio reliably beats. Take any employer match first — an immediate 50–100% return — then attack the debt.
  4. Fill the tax-advantaged space, in order. In Canada that is the TFSA, RRSP and FHSA; elsewhere substitute your equivalents. Know what the public floor gives you — in Canada, CPP and OAS — and plan the gap above it.
  5. Automate, then leave it alone. A standing transfer on payday, a broad low-cost index fund, one review a year.

Two ideas from the framework worth adopting early, because they shape what you accumulate toward:

  • A guaranteed income floor. Identify the non-negotiable baseline — housing, food, utilities, insurance — and aim to cover it with inflation-hedged income you cannot outlive, rather than from portfolio withdrawals. Everything above the floor can then absorb volatility.
  • Dynamic decumulation. Which account you draw from, in which year, is a tax decision worth as much as a percentage point of return. It is also the part most people improvise, and the part software genuinely helps with.

For a target, the common starting heuristic is 25× annual spending — the inverse of the “4% rule”. Treat it as a first approximation: it assumed a roughly 30-year horizon and a particular US portfolio, and OWL’s own premise is that thirty years is the wrong model.

A prompt that gets a useful answer — adapt the numbers:

I am 43, in Canada, and spend about $58,000 a year. I have 8 months of expenses in cash and no high-interest debt. I can save roughly 18% of gross income. Draft a contribution plan across TFSA, RRSP and FHSA for the next five years, state your assumptions, show the arithmetic, and list the five questions I should ask a fee-only planner before I commit.

Then verify every contribution limit against the tax authority’s own site.

Keep the plan alive

Here is the part most people skip, and where this site’s usual habits transfer directly: treat all three plans as documents, not decisions.

Put them in three plain Markdown files. Put them under version control. When something changes — a raise, an injury, a rate change, a birthday — edit the file and commit it. You end up with a dated history of what you believed and why, which is exactly what you want when reviewing a decade of choices, and exactly what an AI needs as context for a useful review.

Some desktop AI tools can be pointed at a local folder and read those files directly, which turns the review below into a single instruction rather than a copy-and-paste exercise. Whether yours can is worth checking; the method does not depend on it.

A quarterly ritual, twenty minutes:

  1. Paste the current plan file into a model.
  2. Ask: what has changed, what assumption is now wrong, what did I not do?
  3. Update the markers — new skills picked up and standing social commitments for the mind; grip strength and sit-to-stand for the body — alongside the portfolio number.
  4. Edit the file. Commit with a message saying what changed and why.

That is the whole system. Reviewing the cognitive and physical markers in the same sitting as the financial ones is the only part that is unusual, and it is the part that makes it OWL rather than three plans in separate drawers.

Sources worth reading directly