Skip to main content
RoastIQBuyerLensHugoPricingBlogAbout
Book a demoSign inStart free →
AI Agent · Autonomous creative intelligence

Meet Hugo

Your creative
strategist that
never sleeps.

Upload a video, ask a question, and Hugo runs the full pipeline — RoastIQ diagnostic, buyer interviews, benchmark search, creative brief — all from one conversation.

Hugo uses RoastIQ + BuyerLens + Benchmarks as tools. Pro plan.

Run RoastIQBuyer InterviewsSearch BenchmarksCompare CreativesGenerate BriefCheck Project
Hugo
Hugo
AI Creative Intelligence Agent
Agent

The shift

Stop navigating screens.
Start asking questions.

The pipeline is already powerful. The cost is the context-switching, uploading here, comparing there, exporting somewhere else. Hugo collapses all of it into a single conversation that ends in a decision.

01

One conversation. Full pipeline.

Diagnostic, buyer interviews, benchmark search, creative brief, Hugo picks the right tool for the question. You never leave the chat.

02

Reasoning you can audit.

Every tool call, every piece of evidence, every decision is shown. You trust the recommendation because you can see exactly how it got there.

03

Built on the SaliencyLab stack.

Not a wrapper. Hugo calls the real RoastIQ scoring engine, the real BuyerLens panels, the real benchmark pool, the same data your team already trusts.

How Hugo thinks

Watch the reasoning, not just the answer.

Hugo shows you every tool he uses, every piece of evidence he reads, every decision he makes — so you trust the recommendation.

Step 01
Understanding
Parses what you’re trying to decide, scale, fix, compare, or kill.
Step 02
Reading report
Pull the RoastIQ diagnostic: 5 KPIs, verdict, benchmark context, evidence.
Step 03
Searching benchmarks
Query the pool for platform + category norms to contextualize the scores.
Step 04
Deliberating
Cross-reference evidence. Identify the single biggest weakness and why it matters.
Step 05
Delivering the answer
One clear recommendation with evidence, not a list of maybes.
Understanding
Hugo reads your question and identifies the decision type: "Why did this ad get Sharpen?" = diagnostic explanation + fix recommendation. He needs: the RoastIQ report, benchmark context, and the specific KPI weakness.
Decision type: Explain verdict + recommend fix
Required data: 5 KPI scores, composite, benchmark position
Tools needed: read_report → search_benchmarks → synthesize

Hugo's toolkit

12 tools. One agent. Zero manual steps.

Hugo decides which tools to use based on your question. You never navigate between screens.

Run RoastIQ
Full 5-KPI diagnostic on any uploaded creative in under 90 seconds.
📊
Read Report
Pull any saved RoastIQ result, scores, verdict, evidence, next move.
👥
Buyer Interviews
Run 1–15 synthetic buyer personas with structured interview spines.
🔍
Search Benchmarks
Query the pool by platform, category, format, and region.
🔬
Compare Creatives
Side-by-side comparison across versions with per-KPI delta.
📋
Generate Brief
Prioritized creative direction: what to fix, keep, and test next.
📤
Upload Asset
Register uploaded files directly from conversation for analysis.
⏱️
Check Job Status
Poll running analyses and interviews, know when results are ready.
📝
Log Decision
Record scale/sharpen/rebuild decisions to the project audit trail.
🎯
Check Project
Full workspace state: assets, reports, pending jobs, decision history.
🗃️
Query Database
Direct read access to all data, reports, benchmarks, attributes.
Audit Trace
Replay any conversation step by step, every tool call, every answer.

Sample answer

Not a wall of text. A decision.

Hugo returns the verdict, the KPI breakdown, the evidence behind it, and a specific next move, formatted for the room, not the model.

Hugo · Diagnostic answer

Coca-Cola · Holidays 2025 · YouTube 60s

Verdict: Sharpen
81
Beat Skip
79
Noticed
82
Brand
63
Sell Proposition
81
Build
Composite 77, above the Sharpen threshold (55–69), but Sell Proposition drags at 63, the only KPI below 70.
Product timing gap. Bottle appears at 0:00, then disappears for 59 seconds. Truck convoy at 0:22, end card at 0:57.
Platform fit. Works on YouTube (viewers reach the end card). Collapses on TikTok and Reels, the skip decision happens at 2–3s.
YouTube FMCG 60s pool (n=211): composite median 68. This ad sits at p72. Sell Proposition norm is 72, this ad sits below p25.
Next move

Add a product moment between seconds 5 and 8, while the emotional hook still carries. On Meta placements, add a text overlay: 91% of viewers will hear no jingle.

What Hugo is worth

One agent replaces three Tuesday meetings.

A planner, a strategist, and a research lead, usually three calendars, three decks, and a week of waiting. Hugo runs the same work in the time it takes to read this paragraph.

90s
From upload to first verdict
12
Tools orchestrated in one chat
Without Hugo
  • 3–5 day turnaround per cut, per agency loop
  • Manual hand-off between strategist, planner, researcher
  • Subjective verdicts. No audit trail when a launch flops
  • $40k–$120k for a single Kantar Link AI study
  • Insight buried in a PDF, three weeks after the spend decision
With Hugo
  • Verdict in 90 seconds. Buyer interviews in five minutes
  • One agent, one chat, one project, every artefact linked
  • Every score traceable to evidence, benchmark, and timestamp
  • Included in Pro. No per-study uplift, no enterprise minimum
  • The decision lives where you make it, before the media buy
Compress the loop

What used to take a week of meetings, diagnostic read, buyer reaction, benchmark check, brief rewrite, happens in one conversation while the editor is still in the timeline.

Protect the spend

A single rejected cut saved before paid amplification pays for Pro for a year. Hugo flags Rebuild verdicts before the media plan is locked.

End the deck cycle

No more 40-slide pre-test PDFs. Hugo's output is a verdict, five KPIs, the evidence, and a next move, shareable in Slack, not a research portal.

Defend the decision

Every verdict logs to a project audit trail, model version, benchmark pool, confidence band, evidence. When the CMO asks why, you have the receipts on hand.

Why teams trust him

Recommendations come with their receipts.

Hugo isn't a black box. Every answer ties back to the same data your team already trusts, and you can inspect every step.

Auditable reasoning trace

Every tool call, every database read, every benchmark query is logged and inspectable. Decisions arrive with their evidence chain attached, never as opaque conclusions.

Same data your team already uses

Hugo calls the real RoastIQ scoring engine, the real BuyerLens panels, the real benchmark pool. He doesn't re-implement the pipeline, he orchestrates it. One source of truth.

Honest about what he knows

Hugo flags low-confidence answers, surfaces benchmark sample sizes, and refuses to invent metrics. Scores are model predictions; synthetic interviews are simulated. He says so.

Questions teams ask first

What Hugo is, and isn't.

Straight answers to the questions every CMO, planner, and head of creative asks before they hand him a brief.

Is Hugo a chatbot, or does he actually do the work?+
He does the work. Hugo isn't a wrapper around an LLM, he calls the same RoastIQ scoring engine, BuyerLens panels and benchmark pool your team uses today. When he says "I ran the diagnostic," he actually ran it and the result is in your project, with the same audit trail as a manual run.
What does he cost? Is there a per-question fee?+
Hugo is included in Pro. No metered tokens, no enterprise minimum, no per-question billing. The only quota that applies is the Pro plan's monthly ad and panel allowance.
How is this different from RoastIQ on its own?+
RoastIQ is the scoring engine. Hugo is the strategist who knows when to run it, what to compare it against, which buyer persona to interview, and what brief to write next. RoastIQ gives you a number. Hugo gives you the decision.
Does he replace my creative strategist or planner?+
No. He replaces the mechanical part of their job, the pulling, comparing, deck-building, benchmark-checking, so they spend their hours on craft and client judgement, not screen-switching. Teams using Hugo report he handles the first-round read on every cut, before a human looks at it.
Can I trust the recommendation? What if he's wrong?+
Every answer ships with its evidence trail, which tool was called, what benchmark pool was queried, what sample size backed it, and a confidence band. RoastIQ scores are validated against public engagement and click-intent outcomes (held-out Spearman ρ +0.30 to +0.32 across TikTok and YouTube). Hugo flags low-confidence answers explicitly and refuses to invent numbers.
What data does Hugo see, and where does it live?+
Only what's in your project workspace, your uploads, your scored reports, your benchmark queries. Nothing is sent to third-party model providers for training. The benchmark pool itself is built from publicly available transparency sources (Meta Ad Library, TikTok Creative Center, Google Ads Transparency), never customer data.
Does he predict sales or ROAS?+
No, and he'll tell you so. SaliencyLab predicts engagement and click intent, leading indicators, validated against public outcomes. We don't claim to predict in-market sales, attributed conversion, or brand recall. Hugo holds the same line in every answer he gives.
How do I get started?+
Upgrade to Pro and Hugo appears in your workspace. Upload a cut, ask him anything, "why did this score that?", "what would a Gen-Z buyer say?", "compare this to my best Q4 ad." He picks the tools. You make the call.

Your creative strategist is ready

Stop navigating screens. Start talking to Hugo.

Upload a video, ask a question, and Hugo runs the full pipeline — all from one conversation.

Hugo