---
title: "Agency Creative Concept Shortlisting with AI Panels | Minds"
canonical_url: "https://getminds.ai/blog/agency-creative-concept-shortlisting-ai-panels"
last_updated: "2026-06-13T15:30:08.067Z"
meta:
  description: "Pre-screen 12 to 20 creative concepts with synthetic ICP panels in 2 hours. Walk into the client meeting with 3 defensible directions, not a guessing game."
  "og:description": "Pre-screen 12 to 20 creative concepts with synthetic ICP panels in 2 hours. Walk into the client meeting with 3 defensible directions, not a guessing game."
  "og:title": "Agency Creative Concept Shortlisting with AI Panels | Minds"
  "twitter:description": "Pre-screen 12 to 20 creative concepts with synthetic ICP panels in 2 hours. Walk into the client meeting with 3 defensible directions, not a guessing game."
  "twitter:title": "Agency Creative Concept Shortlisting with AI Panels | Minds"
---

May 23, 2026·Research·Minds Team

# **Agency Creative Concept Shortlisting with AI Panels**

Pre-screen 12 to 20 creative concepts with synthetic ICP panels in 2 hours. Walk into the client meeting with 3 defensible directions, not a guessing game.

[Try Minds free](https://getminds.ai/?register=true)

The creative-shortlist meeting is the single most expensive 60 minutes in an agency workflow. The creative team comes in with 12 to 20 concepts. The strategy lead has 60 minutes to argue down to 3. The account lead is calculating how to defend the choices to the client. Half the team has strong opinions, the other half is waiting for the meeting to end.

Most of the time, the 3 concepts that go to the client were chosen by the loudest voice in the room. That is not a sustainable competitive advantage for an agency, and it produces a 30 to 50 percent kill rate at the client review. Which means 60 percent of creative work gets shipped to the bin.

In 2026, the leverage move is to run every concept through a synthetic ICP panel before the internal shortlist meeting. The panel evaluates 20 concepts in 2 hours against the actual target audience, surfaces the bottom 6 (kill list), the top 3 (defendable shortlist), and the 5 in the middle that benefit from craft investment. You walk into the meeting with data, not opinions.

## Why this works for creative

Creative evaluation is famously subjective, but the parts that actually matter for shortlisting are not:

1. **Comprehension.** Does the audience get the message in 3 seconds?
2. **Relevance.** Does it feel made for them or for everyone?
3. **Distinctiveness.** Does it stand out from the category convention?
4. **Purchase or consideration intent.** Does it move the brand metric?
5. **Risk signals.** What backfires, alienates, or feels off-brand?

These are exactly the dimensions a calibrated persona panel evaluates reliably. The creative director still owns craft, distinctiveness, and narrative judgment. The panel handles the objective evaluation against the target audience.

## The 2-hour workflow

### Step 1: Build the client-specific ICP panel (30 min, one-time)

For each retainer client, build a panel template once and reuse for every concept review:

- 40 personas matched to the client's primary buyer ICP
- Demographic and firmographic accuracy (age range, geo, role, company size, purchase context)
- Behavioral context (current category brand use, attitudinal cohort, recent category purchases)
- Save as a reusable Custom Audience

For a DTC skincare client, that might be:

- Women 28 to 45
- US metro (NYC, LA, SF, Chicago, ATL, Miami)
- Currently use a mid-tier brand (Sephora regulars, not Walmart shoppers)
- Active on Instagram, follow at least 5 beauty creators
- Spend $80 to $250 per month on skincare

This panel becomes the standing audience for every concept the agency develops for that client.

### Step 2: Format the concepts for panel review (30 min)

Each concept gets:

- Headline (1 to 2 lines)
- Visual description or scamp
- Body copy (if applicable, 2 to 4 lines)
- CTA
- Channel context (print, social, TVC, OOH)

You do not need finished art. The panel evaluates the idea, not the production. A scamped concept and a finished concept produce 95 percent of the same panel signal.

### Step 3: Run the panel (45 min)

Present the concepts in batches of 4 to 6 to avoid panel fatigue. For each concept, ask the panel:

- What is the brand trying to say?
- How does this make you feel?
- Does this feel made for you?
- Compared to the typical ad in this category, how does it stand out?
- Would this make you consider the brand?
- Is there anything that feels off, confusing, or annoying?

The panel returns per-persona reasoning. You see comprehension scores, relevance scores, distinctiveness scores, and intent scores per concept.

### Step 4: Triangulate to a shortlist (15 min)

Tag each concept:

- **Kill (bottom 6):** low comprehension, low relevance, or high risk signal
- **Hold (middle 5):** decent scores but not standout, candidates for craft investment
- **Shortlist (top 3 to 5):** high relevance, high comprehension, high intent or high distinctiveness

The kill list is the time-saving punch. Those concepts do not need a 60- minute shortlist meeting. You and the creative director sign off in 5 minutes.

### Step 5: Walk into the shortlist meeting (0 min)

You enter with 5 to 8 concepts, scored, with panel reasoning attached. The meeting argues craft and narrative direction, not "which of these 20 is worth showing the client." The meeting takes 30 minutes instead of 90.

## How this changes the client relationship

The client meeting changes character. Instead of presenting 3 concepts and hoping, you present 3 concepts with a methodology slide:

"We evaluated 18 directions with a 40-persona panel matched to your ICP. Here are the 3 that scored highest on comprehension, relevance, and intent. Here is the panel composition. Here are the discriminating signals."

The client sees:

- Rigor in the process
- Evidence behind the recommendation
- A clear story for why these 3 and not the other 15

Kill rate at client review drops from 30 to 50 percent to 5 to 15 percent. That is not just better win rate. It is fewer re-briefs, fewer late nights, and a stronger argument for premium retainers.

## Real example: B2B SaaS launch campaign

A growth agency built 16 concepts for a B2B observability launch campaign. Before the panel workflow they would have walked into the shortlist meeting with 16 print-outs and a 90-minute argument.

With the panel:

- 16 concepts evaluated by a 40-persona panel of mid-market data engineers
- 5 concepts killed (low comprehension on the technical hook, or generic category framing)
- 6 concepts held for craft investment (scored well on relevance but needed visual polish)
- 5 concepts shortlisted for client review

The client signed off on 3 of the 5 shortlisted concepts on the first review. Production timeline shortened by 2 weeks. The agency used the panel methodology slide as a closing argument in their next new-business pitch and won the account.

## What patterns the panel surfaces consistently

Across hundreds of agency panel runs, recurring patterns emerge:

**Category-convention concepts under-perform on distinctiveness.** Panels flag the concepts that look like the rest of the category. Creative directors have always known this, but panels make it provable to clients.

**Overly clever wordplay fails comprehension.** If the panel cannot explain the brand message in 1 sentence, the audience cannot either. Cleverness that works in the creative room often fails outside it.

**Bold visual + boring copy is a common failure mode.** Panels rate visuals strongly but flag the copy as not following through. Worth flagging at shortlist so the surviving concepts get balanced craft.

**Demographic mismatch in casting kills relevance fast.** Panels notice when the casting choice does not match the implied user. Easy fix at shortlist, very expensive fix in production.

## What panels do not replace

Be honest about the limits:

- **Award craft.** Panels do not predict whether something wins a Cannes Lion. They predict whether it works for the target audience.
- **Long-term brand effects.** Panels measure response, not 24-month brand health.
- **Cultural moment alignment.** Panels do not see the news cycle the week your ad runs.
- **Production quality decisions.** Visual finish, music choice, voice-over selection still require human craft.

The panel is for the shortlist decision and the comprehension/relevance/ intent triangulation. The art is still the art.

## How to staff this

The pattern that works at most agencies:

- **Senior strategist:** sets up client panel templates (one-time, ~30 min per client)
- **Strategist or junior creative:** runs the panel sessions on each concept batch
- **Creative director:** reviews the panel output before the shortlist meeting
- **Account lead:** uses the panel methodology in client conversations

Total agency time on the workflow: under 4 person-hours per concept batch. Replaces 8 to 12 person-hours of internal shortlist arguments and re-work.

## Pricing implication

Agencies that have adopted this workflow are quietly raising their strategic research line items. The pitch is: "We use synthetic ICP panels to validate every concept before client review. That is why our kill rate is 10 percent not 40 percent. That capability is bundled into your retainer."

That is a defendable premium and a real competitive moat against agencies still running creative on gut feel and politics.

## What to do this week

1. Pick your largest retainer client.
2. Build a 40-persona panel matched to their ICP (30 min).
3. Run the next batch of creative concepts through the panel before the internal shortlist meeting.
4. Time how long the shortlist meeting takes versus your average. Track the client kill rate on the resulting shortlist.

Within 2 cycles you will have proof points to scale the workflow across your entire client roster. The agencies that adopt this in 2026 will be the ones quietly winning new business and growing margins through the rest of the decade.

## **Frequently asked questions**

### **How is this different from copy-testing platforms like Zappi or System1?**

Faster, cheaper, earlier in the workflow. Copy-testing platforms validate near-finished creative ($800 to $3000 per ad, 1 to 3 days turnaround). Synthetic panels evaluate raw concepts, scamps, or even one-line briefs at the shortlist stage ($0 marginal cost, 2 hours). You use panels to narrow 20 to 3, then copy-testing platforms to validate the 3 survivors if the client wants the extra rigor.

### **Will the client still trust a synthetic-panel-screened shortlist?**

Increasingly yes, and the framing matters. Position the panel as 'pre-screening to eliminate the obvious losers' rather than 'final consumer validation.' Most clients prefer to see 3 well-evaluated directions than 12 unfiltered ones. Show the panel methodology, the ICP composition, the discriminating signals. Sophisticated clients see this as quality of process.

### **What if our agency works across many client categories? We can't be expert in each one.**

That is exactly where panels help. You can build a category-specific ICP panel in 15 minutes for a B2B logistics client, run their creative through it, and surface insights that match someone who works in that category daily. Panels close the category-expertise gap for full-service agencies.

### **Can junior creatives use this without director oversight?**

Yes, with a structured template. The panel prompt and ICP composition are the only judgment-heavy parts. Once you set up the panel template for a client, any creative on the team can run concepts through it and get directional output. Senior review then happens on the panel-filtered shortlist, not on 20 raw concepts. That is where the time savings compound.

### **What about creative breakthrough concepts that test poorly but are bold ideas?**

Panels surface this clearly. A concept that scores 'low purchase intent, high attention, high distinctiveness' is exactly the breakthrough archetype, and panels rank those differently from safe-but-bland concepts. Present the discriminating scores, not a single composite rank. The creative director still makes the call on whether to push the bold one. The panel gives you the data to argue for it.