The AI-Assisted Trading Plan: From Prep to Journal
Most traders do not lose because they cannot read a chart. They lose because they do not have a plan they can follow under pressure, and because they never review what they actually did. The chart read is the easy part; the system around it is where the money and the mistakes live.
AI can help build that system. It can turn notes into a consistent one-page plan, check the plan for missing invalidation, structure a journal, and summarize a week of behavior. It cannot choose your bias, size your position or carry the accountability for a trade.
This guide is the planning and journaling pillar of the AI Trading Desk. It covers the one-page plan, the 30-minute pre-market routine, scenario design, the journal fields that matter and the weekly review. It is educational material, not financial advice.
A plan is not a prediction. It is a set of conditional instructions written before emotion arrives.
Contents
- Why Trading Plans Fail
- The One-Page Plan
- The 30-Minute Pre-Market Routine
- Bias And Levels
- Three Scenarios, One Invalidation Each
- The Risk Budget
- The Journal Fields That Matter
- The Weekly Review
- Building A Trading Memory
- Where AI Helps — And Where It Does Not
- Common Mistakes
- Frequently Asked Questions
- Related Guides
- Sources
Why Trading Plans Fail
A plan fails for one of five reasons.
- It is a prediction, not a plan. “I think the market will go up” is not actionable. It has no trigger, no invalidation and no risk.
- It is too long to use. A ten-page document is not read at 09:25. A plan has to fit on one page and be usable in the two minutes before the open.
- It has no risk budget. The plan says what to buy but not how much to lose, so the first adverse move forces an improvised decision.
- It has no no-trade branch. The trader assumes a trade must happen. The market does not owe anyone a setup.
- It is never reviewed. The plan is written, the day happens, and nobody compares the two. Without review, the same mistake is reinvented with a new rationale.
AI cannot fix a plan that is fundamentally a prediction. It can fix the form: it can turn a messy note into a structured plan, check for missing invalidation and flag a risk number that exceeds the limit. The judgment remains with the trader.
The One-Page Plan
A useful plan has seven parts. Keep each one short.
1. Context
- Instrument and contract.
- Session and time zone.
- Date and the time the plan was written.
- The news calendar: scheduled releases, Fed speakers, holidays, contract rolls.
- Any macro context that could invalidate the technical read.
The context section exists to stop the plan from being applied in the wrong conditions. A technical setup on a Fed day is not the same trade as the same setup on a quiet Tuesday.
2. Bias
One line, with the reason. For example:
“Bearish while price is below the 60-minute EMA stack and below VWAP; the last lower high at 29,635 is the invalidation for the down-leg.”
A bias without a reason cannot be reviewed. A bias with a reason can be tested and, if necessary, abandoned.
3. Key Levels
A short table of resistance and support, each with its source and whether it is a zone or a line. Mark the two or three levels that matter most. Do not list twenty levels; a plan with twenty levels has no levels.
4. Three Scenarios
Primary, secondary and no-trade. Each with a trigger, entry, stop, target and invalidation. The no-trade branch is mandatory.
5. Risk Budget
- Risk per trade (for example, 1% of equity).
- Daily loss limit (for example, −2%).
- Position cap per account.
- The number of losses that ends the day.
- The maximum number of trades.
6. Rules For Today
The two or three behaviors that are most at risk today. For example:
- “No market shorts below the day low; wait for a retest.”
- “No new entries 15 minutes around the 08:30 release.”
- “No averaging down; if the first entry is wrong, take the stop.”
7. Journal
The fields you will fill during and after the session. A plan without a journal is a wish; the journal is the evidence that the plan was tested.
A one-page template makes this repeatable. The Daily Trading Plan Template is the companion to this guide.
The 30-Minute Pre-Market Routine
The routine is not about spending more time. It is about spending the same time in the same order, so nothing important is skipped.
| Step | Time | Task |
|---|---|---|
| 1. Macro scan | 5 min | Calendar, yields, dollar, index futures, any overnight headline that explains a gap. |
| 2. Chart capture | 5 min | Higher timeframe and lower timeframe screenshots; check timestamps and data quality. |
| 3. Structure read | 5 min | Four-pass review: structure, character and volume, context, conditionals. |
| 4. Level marking | 5 min | Mark the levels that matter; choose the line in the sand. |
| 5. Scenario writing | 5 min | Three branches with trigger, stop, target and invalidation. |
| 6. Risk check | 3 min | Position size, daily loss headroom, position cap, order review. |
| 7. Journal setup | 2 min | Write the bias rationale, the emotional state and the compliance intention. |
Two rules make the routine work. First, the plan is written before the open, not during the first five minutes. Second, the plan is reviewed against the actual open before any entry. The open can confirm, invalidate or leave the plan unresolved; all three are useful information.
AI can help in steps 2–5 by organizing the screenshots and drafting the structure. It cannot do step 1 for you unless you give it the calendar, and it cannot do step 6 reliably without your account inputs. The routine is a human workflow with AI assistance, not an AI workflow with a human witness.
See The One-Page Daily Trading Plan for the template that follows this routine.
Bias And Levels
A bias is a working hypothesis about the direction of least resistance. It is not a commitment to a trade.
Write the bias in one sentence, with the evidence and the condition that would change it:
“Bearish while price is below the 60-minute EMA stack; a 5-minute close above 29,635 would invalidate the down-leg and put the market back in the range.”
This format has three advantages. It names the evidence, it names the invalidation, and it makes the bias reviewable after the session. If the bias was wrong, you can see whether the evidence was wrong or the interpretation was wrong.
Levels should be treated as zones, not exact lines. A prior session low from one indicator may differ from the same label on another timeframe; print both and treat the gap as a zone. The plan should say which levels are hard edges and which are approximate.
The line in the sand is the one level that defines the day. It should be prominent in the plan, with an explicit “above” and “below” behavior. If the plan has no line in the sand, it has no decision point.
Three Scenarios, One Invalidation Each
A scenario is a conditional statement, not a hope. It has five parts:
- Trigger — what price must do to activate the scenario.
- Entry — where the trade is taken, ideally on a retest or a confirmation, not in the middle of the move.
- Stop — where the trade is wrong, beyond the structure it is defending.
- Targets — the first, second and (optionally) third target, with the R multiple for each.
- Invalidation — the condition that says the scenario is dead, even if the stop has not been hit.
The invalidation is the part most traders omit, and the part AI is best at checking. A scenario without an invalidation is not a scenario; it is a bias with an entry.
Three scenarios are enough:
- Primary — the most likely branch, given the evidence.
- Secondary — the credible alternative, often a failed break or a retest.
- No-trade — the branch where neither boundary is accepted, or the context is unclear.
Probabilities are optional. If you use them, keep them rough (for example, 40/35/25) and do not pretend to precision you do not have. The value is in forcing yourself to consider the alternatives, not in the exact number.
A useful AI prompt is: “Here is my plan. For each scenario, tell me the invalidation and the evidence that would make you abandon it. If a scenario has no invalidation, say so.” The model is a consistency checker, not a forecaster.
The Risk Budget
The risk budget is the part of the plan that protects the account from the trader. It should be decided before the session and not renegotiated during it.
A simple budget:
- Risk per trade: 1% of equity.
- Daily loss limit: −2%.
- Position cap: a hard maximum per account (for example, five MNQ contracts), independent of the 1% calculation.
- Loss limit: two full losses end the day, because three 1% losses exceed the −2% daily limit.
- Trade limit: a maximum number of trades, to prevent overtrading.
- News blackout: no new entries within 15 minutes of a scheduled release.
The arithmetic must be explicit: stop distance × point value × contracts × accounts. If the result exceeds the budget, reduce size or tighten the stop to structure. Never widen the stop to keep the size.
AI can run the arithmetic, but it cannot know the contract value or the account state unless you provide them, and it cannot decide whether the stop is structurally sound. Check the inputs, check the output, and keep the final decision with the trader.
See AI for Trading Discipline for the full risk framework.
The Journal Fields That Matter
A journal that records only P&L is a scoreboard, not a learning tool. The fields below are the minimum for a useful review.
| Field | Why it matters |
|---|---|
| Date, time, instrument, session | Context; patterns often cluster by time and instrument |
| Setup name and scenario | Connects the trade to the plan |
| Entry, stop, target | Shows whether the trade was taken at the planned location |
| Planned R multiple | Shows whether the reward justified the risk before entry |
| Exit and realized R | Separates the outcome from the decision |
| P&L | Useful, but not the only measure |
| Plan compliance (1–10) | The variable the trader controls |
| Emotional state before / during / after | Connects behavior to outcome |
| One line: what would have made the trade better | Turns the trade into a lesson |
| One line: what would have made the decision worse | Makes the counterfactual explicit |
The compliance score is the most important field. A profitable trade that broke the rules is a dangerous data point, because it teaches the wrong lesson. A losing trade that followed the plan is a good data point, because it can be repeated. The journal must be able to tell the difference.
The emotional state can be simple: a number from 1 to 10 plus a word (“calm,” “rushed,” “revenge,” “bored”). The goal is not to judge the emotion; it is to see the correlation between state and behavior.
See The Trading Journal Fields That Matter.
The Weekly Review
The weekly review is where a journal becomes a feedback loop. It should take 30–45 minutes and produce one change for the following week.
A useful structure:
- Aggregate the week. Number of trades, win rate, average R, profit factor, largest loss, largest win.
- Check compliance. Average compliance score, the number of trades below 7/10, and the most common rule broken.
- Cluster the deviations. By time of day, setup, instrument, emotional state, after a win or after a loss.
- Compare plan to behavior. Which scenarios were traded? Which were ignored? Which were invented?
- Choose one change. Not five. One behavior to change next week, with a specific trigger and a way to measure it.
- Write it down. The next plan should include the change as a rule.
AI can help with steps 1–4 by summarizing anonymized journal entries and flagging patterns. It is good at counting, clustering and noticing repeated phrases. It is not good at knowing whether a pattern is causal or coincidental. That is your job.
A useful prompt: “Here are 20 anonymized journal entries. Identify the three most frequent deviations from the stated plan, the emotional states that precede them, and the times of day when they occur. Quote the entries that support each pattern. Do not give advice; just show the evidence.”
The output should be a short list of evidence, not a lecture. The trader decides what to change.
See The Weekly Review: Turning a Log into Lessons.
Building A Trading Memory
A single journal entry is a data point. A year of entries is a memory. The difference is structure.
- Use the same fields every day. If the fields change, the data cannot be compared.
- Use the same setup names. A setup called “pullback” one day and “retest” the next is two different setups in the data.
- Record the plan before the trade, not after. Reconstructing the rationale after the outcome is hindsight, not evidence.
- Keep the screenshots. A journal entry without the chart is a story without a source.
- Review monthly, not just weekly. Some patterns only appear over a longer sample.
AI can help maintain the memory by summarizing and searching, but the discipline of recording is human. The value compounds only if the entries are honest.
Where AI Helps — And Where It Does Not
| Task | AI helps | AI does not help |
|---|---|---|
| Writing the plan | Structuring notes into the standard format | Choosing the bias |
| Checking the plan | Finding missing invalidation, inconsistent risk | Deciding whether the setup is worth taking |
| Pre-market routine | Summarizing charts and calendar | Replacing the trader’s judgment |
| Scenario design | Drafting branches and checking for gaps | Assigning accurate probabilities |
| Risk arithmetic | Running the calculation from inputs | Knowing the contract value or the account state |
| Journaling | Summarizing and clustering entries | Making the entry honest |
| Weekly review | Finding patterns and quoting evidence | Deciding what the pattern means |
| Accountability | Reminding you of the rules you wrote | Enforcing them |
The pattern is consistent: AI is a structuring and checking tool. It is not a deciding tool. The trader who understands that gets the benefit; the trader who forgets it gets a more confident version of the same mistakes.
Common Mistakes
- Writing the plan after the trade. A plan written with hindsight is not a plan.
- Using a plan that is too long. If it does not fit on one page, it will not be used.
- Skipping the no-trade branch. The market does not owe you a setup.
- Using probabilities as precision. 63% is not more useful than “more likely than not.”
- Recording P&L but not compliance. Outcome without process is not learning.
- Reviewing only the losses. The profitable rule-breaks are the most dangerous data.
- Changing five things at once. One change per week is enough.
- Letting AI write the journal. The journal is evidence; it has to be yours.
Frequently Asked Questions
What Should A Daily Trading Plan Include?
Context, bias with a reason, key levels, three scenarios with invalidation, a risk budget, the rules for the day and the journal fields. It should fit on one page.
How Long Should A Pre-Market Routine Take?
Thirty minutes is enough if the process is fixed. The goal is not more analysis; it is the same analysis in the same order, so nothing important is skipped.
Should I Use Probabilities In My Scenarios?
Only if they help you think. Rough probabilities (40/35/25) are useful for ranking scenarios; precise probabilities are false precision. The invalidation matters more than the number.
What Is A Compliance Score?
A 1–10 rating of how closely the trade followed the plan. It separates the quality of the decision from the outcome. A profitable rule-break is a low-compliance trade and should be treated as a warning, not a success.
Can AI Write My Trading Journal?
It can summarize and structure entries you have written, but it should not invent the entry. The journal is evidence of what you did and felt; if AI writes it, it is fiction.
How Often Should I Review My Journal?
Weekly for behavior, monthly for patterns. A weekly review should produce one change; a monthly review should test whether the change worked.
What If I Do Not Have Enough Trades For A Weekly Review?
Review the process instead of the outcome: did you write the plan every day, mark the levels, check the risk and complete the journal? Process metrics work with small samples.
Can AI Tell Me Why I Lost Money?
It can show patterns in your journal — times, setups, emotional states, rule breaks — but it cannot know causation. Use the patterns to form a hypothesis, then test it over the next month.
Related Guides
- The One-Page Daily Trading Plan
- The Trading Journal Fields That Matter
- Trading Journal Template
- The Weekly Review: Turning a Log into Lessons
- AI Chart Analysis: What It Can and Cannot See
- AI for Trading Discipline
- AI Trading Prompt Pack
Sources
- US Commodity Futures Trading Commission, Customer Advisory: AI Won’t Turn Trading Bots into Money Machines (25 January 2024). AI cannot predict the future; guaranteed returns are a fraud red flag.
- Barber, B. M., and Odean, T. (2000), “Trading Is Hazardous to Your Wealth,” Journal of Finance, 55(2), 773–806. Active trading carries a performance penalty after costs.
- Barber, B. M., Lee, Y.-T., Liu, Y.-J., and Odean, T. (2014), “The Cross-Section of Speculator Skill: Evidence from Day Trading,” Journal of Financial Markets, 18, 1–24. Less than 1% of day traders predictably earned positive abnormal returns net of fees.
- Chague, F., De-Losso, R., and Giovannetti, B. (2020), “Day Trading for a Living?” SSRN 3423101. Most persistent day traders lost money.
- NIST, Artificial Intelligence Risk Management Framework (AI RMF 1.0), for the human-review principle.
No statistic in this article is invented. Verify figures against the primary source before republication.
Trading involves substantial risk of loss. This article is educational information, not financial, investment, legal or tax advice, and it is not a personal recommendation or solicitation. AI output can be inaccurate or incomplete; every level and figure must be verified on the chart and in the broker platform. Past performance and backtests do not guarantee future results.
Start With a Quick Brief
Book a 30-minute strategy call. We'll diagnose your challenge and map the right AI-powered skill set for your goals.
Book the Call