The AI Project Cycle

FRAME

Suppose you had to build a working AI project from a blank page. You could not train a model before deciding what problem it solves. You could not judge a model before it exists. Some order is forced on you — and it turns out to be the same order, every time.

An AI project moves through six stages: problem scoping, data acquisition, data exploration, modelling, evaluation, deployment. The cycle, not any single stage, is the unit of work. A weak stage caps every stage built on top of it.

The cycle is not a straight line. A poor result at evaluation sends you back to modelling, sometimes further, to data acquisition. Deployment itself feeds new data and new problems back to scoping — the loop, not the finish line, is what closes the project.

six stages, in order — and it loops
Problem scopingData acquisitionData explorationModelingEvaluationDeployment
Deployment → Problem scoping  results send you back — the loop is the work
you cannot skip ahead: each stage needs the one before it; a weak earlier stage caps every later one

↑ Back to top

Problem Scoping (Stage 1)

CONCEPT

Before an AI project touches a single row of data, one stage comes first. Ask what you would need to know before building anything at all. You would need to know whose problem it is, what exactly is wrong, where it happens, and why it is worth solving.

Those four questions are the 4Ws — WHO, WHAT, WHERE, WHY. Answer all four before a single row of data is gathered. Skip one, and you find out only after collecting the wrong data.

A vague problem cannot be solved by a precise machine. The sharpness of the scope sets the ceiling for the whole project. No later stage, however careful, can recover what this first one left undefined.

scope the problem with four questions
WHO — whose problem is it? (the stakeholder)
WHAT — what exactly is wrong? (the gap)
WHERE — in what context does it occur?
WHY — why does solving it matter?
answer all four before gathering data — they decide what data you even need

↑ Back to top

Getting and seeing data (Stages 2-3)

CONCEPT

Stage 2 begins once the problem is scoped. Now you gather data that actually fits it. Ask what data this problem needs. Find sources that hold it. Check each source is reliable. Collect only what is relevant.

More data is not the goal here. The right data beats more data chosen at random. A team can gather whatever is easiest to find, instead of what the scoped problem requires. That shortcut costs it at every stage that follows.

Remember the old rule: garbage in, garbage out. A model can only be as good as the data it learns from — no later stage can repair data that was wrong from the start.

from the question to the right data
Ask: what data does this problem need?Find sources that hold itCheck each source is reliableCollect only what is relevant
garbage in, garbage out — a model can only be as good as the data it learns from
CONCEPT

Stage 3 comes before modelling, never after. Look at the collected data first: chart it, summarise it, and go hunting — for its patterns, its gaps, its surprises.

A problem you see now is cheap; the same problem found after modelling is expensive. Exploration turns a raw table into understanding. It catches the errors and holes that would otherwise be baked straight into the model, unnoticed until it fails.

look before you build
Chart it — make the numbers visible
Find patterns — what tends to go with what?
Find gaps — what is missing or skewed?
Find surprises — what does not fit?
a problem you see now is cheap; the same problem found after modeling is expensive

↑ Back to top

Modeling (Stage 4)

KEY-TERM

Stage 4 is where the decision-making system gets built. The first choice is not which algorithm — it is a simpler question. Is the rule that governs this decision already known and stable?

If yes, write the rule directly: rule-based. If the rule is hidden inside the data, and too complex to state in words, a learning-based model has to infer it from examples instead.

The choice follows the problem, not fashion. Neither path is the advanced one — each fits a different kind of problem, and picking the wrong one wastes the whole stage.

do you already know the rule?
A decision to automate

TEST — Is the rule known and stable?
Yes — rule is known
→ rule-based: write the rule directly
No — rule is hidden in the data
→ learning-based: train a model on examples
CONCEPT

Class 8 named the two paths. Class 9 asks the sharper question: how do you actually choose? Not by instinct — by weighing four factors together.

How complex is the pattern — can an expert state it in words, or not? How much labelled data do you have? How costly is a wrong answer, and must the logic be checkable by an inspector? And must the decision be explainable to the person who acts on it?

Rule-based fits a simple, statable pattern that must be auditable. Learning-based fits a pattern too complex to write down, where enough labelled examples exist to learn it from. The diabetic retinopathy (DR) screening tool shows how the four combine, not just two: no expert can write, in words, the exact retinal signs of early damage (complexity). And enough labelled retinal images already exist to learn the pattern from (labelled data) — both point to learning-based.

Cost and explainability point the other way, toward rule-based — not toward learning-based. A missed case costs real vision, and the decision must be explainable to the person who acts on it: both factors argue for the simpler, auditable path, even though complexity and labelled data argued for learning-based just above.

So the team chooses learning-based anyway, and builds the missing safeguard in: every flagged scan still goes to an ophthalmologist before anyone acts on it, and that check supplies the explainability a bare model score cannot. Weighed honestly, the four factors are not unanimous — two pull each way, and the audit step is what makes the learning-based pick acceptable.

Weigh the four honestly and the choice stops being a guess. The common slip is assuming learning-based always wins because it is more powerful — it fails with too little labelled data, or a decision no one can explain.

four factors decide — not taste
Points to RULE-BASED
Points to LEARNING-BASED
PATTERN COMPLEXITY — simple, stateable in words → rule · too complex to state → learning
LABELLED DATA — little or none → rule · enough to learn from → learning
COST OF A WRONG ANSWER — high, and the logic must be audited → rule · tolerable → learning
EXPLAINABILITY — the actor must see WHY → rule · a correct answer alone is enough → learning
the common slip: 'learning-based is always better, it is more powerful.' It fails with too little labelled data, when the decision cannot be explained, or when a written rule is simpler to audit — weigh all four, not power alone
the DR (diabetic retinopathy) screening tool goes learning-based because no expert can write the retinal-image rule by hand

↑ Back to top

Evaluation (Stage 5) — judging honestly

CONCEPT

Stage 5 is where you judge the model honestly, and honesty here is harder than it sounds. A single accuracy number can hide real failure.

A model can score well overall and still be wrong in exactly the cases that matter most. A single accuracy number can hide failure — "looks good" and "is good" are not the same claim.

Honest evaluation tests on data the model never saw during training. It also looks at which kind of error the model makes — not only how many errors, but which ones, and what each one costs.

one number can lie
SURFACE
Looks goodone high accuracy score on the data it learned from — flattering and unsafe
test on held-out data; read the error types
BENEATH
Judged honestlytested on unseen data, broken down by which errors it makes and what they cost
ask not only 'how often right?' but 'wrong in which way, and who does that hurt?'
KEY-TERM

It feels natural to sum a model up in one number: how often is it right? That single number hides more than it reveals.

A prediction has not one result but four. Two are correct — a true positive, rightly saying yes, and a true negative, rightly saying no. Two are wrong — a false positive, a false alarm, and a false negative, a genuine miss.

The two wrong outcomes are not equal. A missed disease and a false alarm cost entirely different things. That is exactly why one accuracy figure is not enough to judge a model — it treats both mistakes as the same mistake.

four outcomes, not one score
WHAT WAS TRUE →
WHAT THE MODEL SAID ↓
Actually YES
Actually NO
Said YES → correct hit (true positive)
Said YES → false alarm (false positive)
Said NO → a MISS (false negative)
Said NO → correct all-clear (true negative)
a miss (false negative) and a false alarm cost different things — pick the model whose errors you can live with

ai-project-cycle-build.four-outcomes

/50-system/standards/book-look/cell1-a4/adv-sushruta.png

Sushruta

tailorbird

Sort before you judge — one accuracy figure hides four different outcomes, and until each sits in its own clean bin you do not yet know what kind of mistake you are making.

CONCEPT

An honest score needs one more safeguard. Testing on the same data measures memory, not skill — a model tested on the examples it learned from will always look better than it truly is.

The fix is a two-step recipe. Split the dataset before training begins. Train on the larger part. Keep a held-out part completely untouched, and use it only once, at testing time.

Only a score measured on data the model has never seen tells you how it performs on the world it will actually meet.

memory is not skill — split the data
COMPOSED OF ONE DATASET, SPLIT
TRAIN — the model learns from this part
TEST — held out, unseen until the very end
a score on the TRAIN part flatters — the model saw the answers; only the held-out TEST part measures real skill on new cases

↑ Back to top

A model does not know its own limits

CONCEPT

Take a mango sorter, trained on Alphonso mangoes photographed in afternoon light. Feed it Dasheri mangoes under a different bulb. Does it pause, and say, "wait — I do not recognise these"?

It does not. It keeps sorting, handing back a confident YES or NO on every single fruit — even though those answers can no longer be trusted.

A model has no inner voice. It cannot tell you when it has left the conditions it was trained and tested on. It just keeps answering, exactly as sure of itself as before, while its real accuracy quietly falls apart underneath the confidence.

That blind confidence, with nothing inside to raise a flag, is the AI blindspot. It is exactly why a deployed model must be watched from the outside — it will never warn you itself.

the model cannot flag its own blindspot
SURFACE
Inside the training zoneAlphonso mangoes, afternoon light — confident answers that are trustworthy
the model has no inner voice — it cannot know it left the zone
BENEATH
Outside the training zoneDasheri mangoes, a different bulb — the same confident answers, now unreliable, no warning raised
this is WHY deployment needs outside monitoring — a model will never flag its own blindspot

↑ Back to top

Deployment (Stage 6)

CONCEPT

Stage 6 puts the model into real use, and watches it there. Shipping the model is not the end of the work — it is the start of a different kind of work.

Monitor whether it still performs as well in the real world as it did in testing. Watch for drift, as the conditions around it change in ways the training data never showed it.

Deployment is where the cycle closes and restarts — not where it ends. New data and new problems it surfaces feed straight back to Problem Scoping.

ship, watch, loop back
Put the model into real useMonitor: is it still right in the world?Watch for drift as conditions changeFeed new data + problems back to scoping
deployment is where the cycle closes and restarts — not where it ends
KEY-TERM

It is tempting to think the project ends once the model is deployed and working. It does not. Deployment opens three ongoing jobs, not one act of shipping.

Access: put the tool in a form real users can actually reach — a tablet app, a printout, an API — with no expert needed on site. Monitoring: check regularly whether the scoreboard is drifting from its launch numbers, against an agreed threshold that triggers a flag-and-retrain.

Limits communicated: the tool's evaluation limits must reach every user who acts on its output, printed on every result — never buried in a manual nobody reads.

The work starts at shipping and ends only at retirement. A model shipped and then forgotten drifts quietly into danger, with nobody watching.

three jobs, from launch to retirement
ACCESS — put it in a form real users can reach (app, printout, API); no expert required on site
MONITORING — check the scoreboard against launch numbers; cross the agreed threshold → flag and retrain
LIMITS COMMUNICATED — every user who acts on an output sees the tool's limit, on every result, never buried
the work starts at shipping and ends only at retirement
the common slip: 'once it is deployed and working, the project is done.' Deployment STARTS the monitoring phase — ship-and-forget lets a once-good model drift into danger without warning
The AI Project Cycle — Grade 9 (417) · projected from the LATTICE via prism_html.py · register: school-g9

↑ Back to top