The strategic case for Layer 3 workflows, made in The Three Layers of AI Fluency, is easy to make at the abstract level. Layer 1 is prompts, Layer 2 is patterns, Layer 3 is workflows. The compounding lives at the top. Everyone nods. Then almost everyone goes back to opening blank chat windows because they have no concrete picture of what a Layer 3 workflow actually looks like for their kind of work.
This piece fixes that. Five workflows that real operators are running in 2026, broken down to the components: the trigger, the context, the templates, the output, the iteration. None of them are hypothetical. None of them are about exotic use cases. They are the boring, high leverage workflows that compound week after week because the operator did the upfront work of building them and keeps editing them as they learn. Read all five. Steal the one closest to your role.
Workflow 1: The Sunday Evening 1 on 1 Prep
The operator. A senior manager with six direct reports.
The trigger. Calendar based. Every Sunday at 6pm. The workflow runs whether the manager remembers or not.
The context. A Claude Project loaded with each direct report's job description, current goals, the last three 1 on 1 notes, recent feedback exchanges, and the manager's standing observations about each person's development priorities. The context document is around 600 words per direct report, updated monthly.
The templates. Two patterns run in sequence. First, the per person agenda template, which takes the current week's calendar for that direct report, recent shared documents, and the standing context, and produces a three to five topic agenda. Second, the cross team scan template, which compares the six agendas and flags which direct report's week probably needs the most attention from the manager.
The output. A single Notion page titled "Week of [date] 1 on 1 agendas." Each direct report has a section. The cross team scan sits at the top. The manager opens the page Monday morning and spends ten minutes editing before the day starts.
The iteration. Once a month, the manager edits the per person context documents based on what came up in the actual 1 on 1s. Once a quarter, the manager edits the agenda template itself, removing what was not being used and adding what kept being needed manually.
The compounding. Month one, the workflow saves about an hour of Sunday evening prep. Month six, the workflow has produced 1 on 1s that are noticeably sharper than peers who do not prep this way. Direct reports stay longer. Performance issues surface earlier. The compounding is not in the time saved. It is in the quality of the manager job that becomes sustainable.
Workflow 2: The Candidate Evaluation Synthesizer
The operator. A hiring manager running multiple loops per quarter.
The trigger. Event based. When a candidate completes the interview loop, the workflow runs automatically.
The context. A custom GPT loaded with the role's job description, the scorecard rubric, the company's hiring principles, and the calibration committee's known biases (e.g. "we tend to over weight pedigree on tie cases, push back if the case rests on pedigree").
The templates. Three patterns. First, the interview synthesizer takes all written feedback from the loop and produces a structured summary against the scorecard. Second, the gap finder identifies where interviewers disagreed and what each strong opinion was based on. Third, the calibration prep produces the one page memo the manager brings to the calibration meeting, including the recommended hire decision, the strongest evidence for it, and the rebuttal to the most likely objection.
The output. A Notion page in the candidate's record, attached to the ATS, titled "[Candidate name] hire memo." The output is ready 30 minutes after the last interviewer submits feedback.
The iteration. After each calibration meeting, the manager notes which parts of the memo did the heaviest lifting and which were ignored. The template gets edited monthly. Six months in, the calibration prep is shorter and sharper than it was at the start because the noise has been edited out.
The compounding. Month one saves two hours per hire on synthesis. Month six produces calibration meetings that are noticeably more decisive because the memo does the structural work that previously fell to debate.
Workflow 3: The Board Memo Drafter
The operator. A founder running quarterly board cycles.
The trigger. Calendar based. Two weeks before each board meeting.
The context. A Claude Project loaded with the last four board memos, the board's standing questions and concerns, the company's current strategic narrative, the metrics dashboard, and the founder's preferred tone ("direct, no marketing language, name the problems before the solutions").
The templates. Four patterns run in sequence. First, the metrics narrative template takes the current quarter's numbers and drafts the section that interprets the variances. Second, the strategic update template takes the major decisions of the quarter and drafts the section explaining each. Third, the risk surfacing template generates the section the founder is most tempted to soften and pressure tests it for honesty. Fourth, the question anticipation template predicts the questions the board will ask and drafts the answers the founder will need.
The output. A Google Doc titled "[Quarter] board memo draft." The founder spends half a day editing in their own voice. The full board memo cycle that used to take a week now takes a day and a half.
The iteration. After each board meeting, the founder notes which questions the board actually asked and which the workflow correctly predicted. The question anticipation template gets edited. The board memo template gets refined based on which sections the board engaged with versus skimmed.
The compounding. The first quarterly cycle saves about three days. By the fourth cycle, the workflow produces a draft that is closer to the final board memo than the founder's previous first drafts ever were, because the workflow has been tuned to the specific board.
Workflow 4: The Decision Audit
The operator. An executive who wanted a real retrospective practice and never managed to sustain one.
The trigger. Calendar based. The first Sunday of each month.
The context. A Claude Project loaded with the executive's decision log (a simple Notion database where they record each significant decision with date, context, expected outcome, and stake), their calendar for the month, and the company's recent results.
The templates. Two patterns. First, the decision audit template reviews each decision from the prior month, scores it on whether the outcome matched the expectation, identifies which decisions look better in hindsight and which look worse, and surfaces the pattern (e.g. "decisions made under time pressure are systematically worse than decisions made with one extra day of thought"). Second, the next month framing template generates the one page reflection the executive will actually read, focused on the decision they need to make differently this month based on what the audit found.
The output. A one page reflection in the executive's Notion homepage, dated, archived in a Decision Audits folder. The executive reads it at the start of each month.
The iteration. The decision log gets edited as the executive notices what is missing from it. The audit template gets sharper as the executive learns which categories of decisions tend to go wrong in their specific role.
The compounding. Month one produces a useful but ordinary retrospective. Month six produces a reflection practice that has noticeably changed how the executive makes decisions, because the patterns have surfaced repeatedly enough to act on. This is the workflow that most operators try and abandon. The ones who sustain it are unrecognisable a year later.
Workflow 5: The Competitive Intelligence Brief
The operator. A strategy lead inside a mid sized company.
The trigger. Calendar based. Every Tuesday at 7am.
The context. A Gemini Gem (using the deep research mode) loaded with the defined set of competitors, adjacent companies, and industry analysts the strategy lead tracks. Each entity has a profile noting what to watch for (product launches, funding rounds, executive moves, public statements).
The templates. Three patterns. First, the news scan template pulls the last week of news, funding announcements, and public statements from the tracked set. Second, the signal extractor identifies what genuinely changed versus what was noise. Third, the strategic implication template translates the changes into questions to bring to the next strategy meeting.
The output. A one page brief in the strategy team's shared Notion, titled "Week of [date] competitive scan." The strategy lead reads it on the train Tuesday morning. The strategy meeting on Wednesday no longer opens with people sharing news that should have been background.
The iteration. The tracked set gets refined quarterly. The signal extractor template gets edited as the strategy lead notices which signals tend to matter and which tend to be noise. By month six, the brief is producing the questions that the strategy team is asking before they ask them.
The compounding. The first brief takes effort to make useful. The tenth is producing intelligence that the strategy team did not previously have. The thirtieth is the reason the company saw a major competitor move two weeks before peers did.
The pattern across all five
Notice what is the same about these workflows and what is different.
What is the same: every workflow has a trigger that is not the operator's memory. Every workflow has persistent context that lives outside the chat window. Every workflow has templates that have been edited multiple times. Every workflow has an output destination that is not a chat tab. Every workflow has an iteration loop that the operator actually runs.
What is different: the role, the cadence, the specific work, the model of choice, the tool stack. The workflows are not interchangeable. They are tuned to the specific job, the specific operator, the specific company.
The takeaway, for anyone designing their own Layer 3 workflows, is that the components are universal and the contents are personal. You cannot copy someone else's workflow and run it. You can copy the structure (trigger, context, templates, output, iteration) and fill it with your own work. That is the skill the rest of this decade will reward.
What to steal first
If you are deciding which workflow to build first, the honest selection criteria is: which piece of work do you do every week, costs you more than an hour each time, and produces output that someone else relies on?
For most managers, that is the 1 on 1 prep workflow. For most hiring managers in growth mode, that is the candidate evaluation workflow. For most founders, that is some version of the board memo or investor update workflow. For most executives who wanted a reflection practice, that is the decision audit. For most strategy or product roles, that is some version of the competitive scan.
Pick one. Build the minimum viable version. Run it for a month. Iterate. Build the second one after the first has compounded enough that the second feels obvious.
Where to go next
For the strategic frame these workflows fit inside, read The Three Layers of AI Fluency.
For the diagnostic that tells you where you currently sit on the stack and which workflow to build first, run the AI Workflow Audit.
For the protocol to actually build the workspace these workflows run on top of, follow The First Saturday.
For Layer 2 pattern libraries to load into your workflows, the PromptLeadz Free Vault frameworks are the fastest path: SHAPE for the manager workflows, HIRED for the candidate workflows, CRITIC for the decision audit workflow, and the rest of the vault for the others.
For the Pro Pack workflow configurations that wire the patterns directly into Claude Projects and Custom GPTs ready to run, the PromptLeadz Pro Collection is at $29 per pack.
PromptLeadz publishes battle tested AI prompt packs for founders, product, sales, marketing, operations, HR, finance, customer success, adversarial thinking, hard conversations, new role launches, job searches, money conversations, office politics, and managers. All prompts are LLM agnostic. Pricing is in USD.
Laisser un commentaire :