Smart MarketerSMART MARKETER LIVE

SMART MARKETER LIVE 2026 · DAY 1 TAKEAWAYS

AI Creative Workflows: Research, Focused Skills and Human Feedback

Alex Cooper

Updated 2026-10-27 · 1 processed session

Turn creative research and repeated production tasks into a process your team can improve. Build focused AI skills, keep human judgment in charge and use specific feedback to make the next output more useful.

01 / Alex CooperAI Creative Workflows: Research, Focused Skills and Human Feedback

2026-10-20 · AI creative workflows · Ad research · Reusable skills · Human feedback · Creative operations

Turn creative research and repeated production tasks into a process your team can improve. Build focused AI skills, keep human judgment in charge and use specific feedback to make the next output more useful.

PUT THIS SESSION TO WORK

Welcome. Alex Cooper’s roundtable discussion shows how research, focused AI skills and expert feedback can support a better creative process.

Bring together account performance, customer language and creative research, then evaluate proposals against the criteria that matter.

01

Judge ideas against evidence and a clear rubric.

Bring together account performance, customer language and creative research, then evaluate proposals against the criteria that matter.

02

Keep the strategist’s judgment in the process.

Understand the customer research, select the worthwhile ideas and explain the reasons behind creative decisions.

03

Build focused skills for repeatable creative tasks.

Use approved examples and task-specific context for each format instead of relying on one oversized instruction.

04

Improve the workflow with specific expert feedback.

Have qualified practitioners review outputs, identify issues such as pacing or clip selection and refine the skill.

05

Set boundaries for shared skills and decisions.

Define human review points, control who can change instructions and regularly resolve conflicting guidance.

Which useful next step would you like to work on?

Open a choice to see the first question and what you’ll create. Continue the work in your connected Coach.

Design my ad-idea rubricA simple evaluation rubric and a shortlist of ideas to review.

A simple evaluation rubric and a shortlist of ideas to review.

Useful inputs

  • Approved Offer Profile and Persona Profile
  • Relevant business information the user chooses to share

Work through it

  1. Use the approved offer and persona to keep this activity relevant to the selected business.
  2. Identify available account evidence, customer language and creative references.
  3. Define a small set of criteria and explain any user-approved weights.
  4. Evaluate a few supplied ideas, keep uncertainty visible and let the user choose what proceeds.

If inputs are missing: Ask one question at a time for missing information. Mark unknowns clearly; do not invent business facts, results or benchmarks.

Your working document

  • Available evidence sources and unknowns
  • Evaluation criteria and agreed weights
  • Idea-by-idea assessment with reasons
  • User-selected shortlist
  • Next research or production step

Complete when: The user has reviewed the output, confirmed a practical next step and assigned an owner.

Design one focused creative skillA reusable instruction draft for one creative format, with inputs and review criteria.

A reusable instruction draft for one creative format, with inputs and review criteria.

Useful inputs

  • Approved Offer Profile and Persona Profile
  • Relevant business information the user chooses to share

Work through it

  1. Use the approved offer and persona to keep this activity relevant to the selected business.
  2. Choose a narrow task and approved examples.
  3. Document inputs, intended output and the decisions an expert normally makes.
  4. Draft the skill and a small test plan with a human quality review; do not assume the host can render video.

If inputs are missing: Ask one question at a time for missing information. Mark unknowns clearly; do not invent business facts, results or benchmarks.

Your working document

  • One selected task and intended output
  • Required inputs and approved examples
  • Reusable instruction draft
  • Quality criteria and human reviewer
  • Small test, feedback plan and host limitations

Complete when: The user has reviewed the output, confirmed a practical next step and assigned an owner.

Build my creative feedback processA review loop defining who gives feedback, what they assess and how changes are approved.

A review loop defining who gives feedback, what they assess and how changes are approved.

Useful inputs

  • Approved Offer Profile and Persona Profile
  • Relevant business information the user chooses to share

Work through it

  1. Use the approved offer and persona to keep this activity relevant to the selected business.
  2. Choose the skill owner and qualified reviewers.
  3. Create a feedback template for concrete issues and explanations.
  4. Define how approved changes are recorded and how conflicting feedback is flagged for resolution.

If inputs are missing: Ask one question at a time for missing information. Mark unknowns clearly; do not invent business facts, results or benchmarks.

Your working document

  • Skill owner and qualified reviewers
  • Feedback template: issue, reason and suggested change
  • Who may propose and approve edits
  • Record of approved changes
  • Conflict review and escalation process
  • Next review date

Complete when: The user has reviewed the output, confirmed a practical next step and assigned an owner.

These activities are Coach-designed applications of the reviewed teaching. Before personalizing them, reuse approved business context or gather the business name and website, draft Offer and Persona Profiles, and obtain approval of both. Summaries do not require onboarding. Coaching paths and output templates are Coach-designed applications. This guide does not submit or store answers.

Read all takeaways ↓ · You can also ask the Coach a question about the session.

Questions about the session

Should one general skill handle all creative formats?

Alex prefers focused skills with examples and context for the specific format or task. He describes moving from a general editing instruction toward separate styles.

Talk · 10:55; Talk · 22:19

Who should give feedback to shared skills?

People with the relevant expertise should review and edit them. The session recommends permission boundaries and a regular review for conflicting guidance.

Talk · 23:00; Talk · 23:50

Can the AI choose every ad idea on its own?

The proposed process keeps final selection with a strategist who understands the research and gives reasons for the choices.

Talk · 19:43; Talk · 21:34

Explore the full session index · 21 entries

Index of reviewed teaching from the supplied recordings. Repeated rotations are consolidated; attendee exchanges and private details are excluded.

An evidence-based idea rubric

Evaluate ad ideas against a defined rubric using evidence from the ad account, customer voice and organic content. Adjust the emphasis to the type of idea you want to develop.

Talk · 01:12

Capture the reasons behind feedback

Have strategists explain which proposed ideas they like or reject and why. Capture that feedback in the working knowledge base so it can inform later suggestions.

Talk · 01:56

Move selected ideas into production

Use research to produce candidate ideas, have a strategist select the useful ones, and move selected ideas into the briefing and scripting process.

Talk · 02:51

Create skills for specific edit formats

Build focused editing skills for different formats rather than assuming one general video-editing instruction will serve every style equally well.

Talk · 10:55

Compare edits and refine the skill

Compare AI-produced edits with examples the team already considers good. Ask experienced editors for specific feedback and refine the skill through repeated review.

Talk · 12:19

Make editing feedback concrete

Make edit feedback concrete: review pacing, clip choice and B-roll selection rather than only saying an output is good or bad.

Talk · 13:12

Involve experienced practitioners

An AI champion can build the first version, then involve experienced practitioners in testing and improving it.

Talk · 13:56

Use focused skills and relevant context

Use small skills with context specific to the format or task, such as a particular edit style or script type.

Talk · 22:19

Look beyond the immediate niche

Creative inspiration can come from within a niche and from other industries. Judge its relevance rather than assuming industry similarity is the only useful criterion.

Talk · 25:20

Document creative decisions

Document the reasons behind creative choices, even retrospectively, so the AI workflow has access to the human decision process.

Talk · 26:38

Set creator-sourcing boundaries

For creator-sourcing workflows, define selection criteria and budget boundaries, including cases that must be flagged for human review.

Talk · 05:32

Keep human review in creator selection

Keep a human review point in creator selection and capture feedback when a suggested creator does not fit the brand or intended audience.

Talk · 06:12

Start with a script, footage and examples

Start an editing experiment with the intended script and footage. Once a useful approach exists, use approved example edits to describe the style in a reusable skill.

Talk · 13:21

Recognize the limits of simple editing examples

The editing examples discussed are relatively simple ad edits, such as cutting footage and adding captions; they are not proof that every complex production task is solved.

Talk · 15:36

Build a recurring review loop

Add a feedback loop to a reusable skill: review each output, explain specific issues and refine the instructions with qualified teammates.

Talk · 16:15

Understand the research personally

On a new brand, do enough research yourself to understand its customers and selling context. A generated report is more useful when the strategist understands the underlying material.

Talk · 19:43

Build a recurring research pack

A recurring research pack can combine ad-account evidence, customer reviews, comments, surveys, community research and competitor observations into proposed ad ideas.

Talk · 20:41

Keep final idea selection with the strategist

The strategist keeps the final selection decision. Explain why selected ideas are useful and why others are rejected so future suggestions have better guidance.

Talk · 21:34

Control who can change shared skills

Control who can change shared skills. Assign feedback and editing rights to people with the relevant expertise rather than allowing conflicting advice from everyone.

Talk · 23:00

Review conflicting instructions

Periodically review the knowledge base for conflicting instructions or feedback. Flag issues for a person to resolve rather than silently deleting them.

Talk · 23:50

Learn through practical experiments

Learning an AI workflow takes practical experimentation. Expect some experiments to be unhelpful and choose an amount of time you can reasonably invest.

Talk · 03:36

Key takeaways

  1. Judge ideas against evidence and a clear rubric.

    Bring together account performance, customer language and creative research, then evaluate proposals against the criteria that matter.

    Talk · 01:12; Talk · 20:41

  2. Keep the strategist’s judgment in the process.

    Understand the customer research, select the worthwhile ideas and explain the reasons behind creative decisions.

    Talk · 02:51; Talk · 26:38; Talk · 19:43; Talk · 21:34

  3. Build focused skills for repeatable creative tasks.

    Use approved examples and task-specific context for each format instead of relying on one oversized instruction.

    Talk · 10:55; Talk · 22:19; Talk · 13:21; Talk · 15:36

  4. Improve the workflow with specific expert feedback.

    Have qualified practitioners review outputs, identify issues such as pacing or clip selection and refine the skill.

    Talk · 01:56; Talk · 12:19; Talk · 13:12; Talk · 13:56; Talk · 16:15

  5. Set boundaries for shared skills and decisions.

    Define human review points, control who can change instructions and regularly resolve conflicting guidance.

    Talk · 05:32; Talk · 06:12; Talk · 23:00; Talk · 23:50

Context to keep in mind

  • The recordings start during Q&A and include two conversations; this entry summarizes retained expert guidance rather than claiming a complete prepared talk.
  • Named model capabilities and product features were not independently verified. These activities design a process; they do not promise automated video rendering or unsupervised outreach.
  • No new resource link has been verified from these transcripts.