Hawky × Abbott

The two checkpoints

Catch it in the script. Confirm it in the video.

Two passes, each one AI first and a trained human second. The order matters: a script fixed before shooting costs nothing, and a video that never had to be reshot is the whole saving.

01Script reviewCheckpoint one

Does the script match the brief, the samples, and MLR?

Before the influencer shoots anything.

Checked against

The approved brief and sample scripts — is this the same claim territory, the same structure, the same product framing. Then Abbott's weighted medical and regulatory criteria, one by one.

Returned as

A score, a verdict, and for anything red the exact criterion that failed and the line that triggered it — so the influencer knows what to change rather than that they failed.

02Video reviewCheckpoint two

Does the finished video match the approved script?

A cleared script does not guarantee a clean video.

Script fidelity

What is actually said against what was approved. Influencers improvise, drop qualifiers, and add lines that were never in the script — this is where that surfaces.

Visual compliance

On-screen text and supers, product depiction, who is on camera and in what role, language and code-mixing actually spoken, and anything implied visually that was never said aloud.

Why both

Script review is where the cost saving is — a rejected script costs a rewrite, a rejected video costs a reshoot and a creator relationship. Video review is where the risk is, because what was approved and what was published are not the same document.

How the scoring works

Your criteria, your weights, your thresholds.

Abbott supplies the rulebook: ten medical and ten regulatory criteria, each weighted by importance. Hawky encodes them and applies them identically to every asset. We are not writing the standard — we are enforcing the one you set.

A high weight is a veto, not a heavier number. Seventy percent of criteria met clears an asset — unless a high-weight criterion failed, in which case it goes back regardless of how well it scored elsewhere. That single rule is what keeps a strong overall score from carrying a serious breach past the gate.

Step 1 · Decompose

Each guideline becomes a testable criterion

A guideline written for a human reader is turned into something a system can evaluate: what it tests, which surface it applies to, and what specifically constitutes a failure. Your MLR reviewer confirms each definition before it goes live — we do not interpret a rule on your behalf.

Step 2 · Weight and classify

You set the weight and the guardrail status

Every criterion carries the weight Abbott assigns it. Separately, you mark which criteria are guardrails — the high-priority ones that trigger red on their own. Weight and guardrail status are two different settings, because a criterion can be heavy without being absolute.

Step 3 · Evaluate and aggregate

Scored per criterion, then rolled up

Each criterion is evaluated independently against every surface of the asset, returning pass, fail or escalate with its evidence. The weighted roll-up produces the percentage; the guardrail check runs separately and overrides it.

Criterion	Type	Surfaces read	Weight	Fails when
Approved claim set	Guardrail	Spoken, on-screen	Veto	A benefit claim outside the approved list appears on any surface
Mandatory disclosure	Guardrail	On-screen, spoken, caption	Veto	Partnership disclosure absent, or below the required prominence or duration
Substantiation	Scored	Spoken, on-screen	High	A statistic or study reference stated that does not match an approved claim
Comparative language	Scored	Spoken, on-screen	High	Product compared to a competitor or category without substantiation
Terminology	Scored	All surfaces	Medium	Unapproved product descriptor or condition name used
Audience framing	Scored	Visual, spoken	Medium	Product depicted for an audience outside the approved indication

Illustrative structure only — the criteria, their weights and their guardrail status come entirely from Abbott

Calibration and change control

Before go-live the encoded set is run against assets your team has already passed and failed, and the weights are adjusted where our verdicts and yours disagree. After go-live, guidelines change under version control: every verdict records the rulebook version it was judged against, so content approved under one version is never silently assumed to pass under the next.

Green flag

Passes to Abbott MLR

Seventy percent or more of weighted criteria met.
No high-weight criterion failed — this condition is absolute.
Reaches your MLR team already checked, with the full criterion-by-criterion record attached, so final sign-off is a confirmation rather than a first read.

Your reviewers spend their time confirming, not searching.

Red flag

Back for re-editing

Below the threshold, or any high-weight criterion failed.
Returned to the influencer with the specific criteria that failed and the line or frame that triggered each one.
Re-submitted content is re-scored, so the fix loop closes without a human re-reading from scratch.

Never reaches your MLR team — which is the point.

The process

Ten steps, upload to audit trail.

Two checkpoints, each AI first and human QA second, with a defined escalation path at both.

01
Abbott uploads campaign inputs
- Brand brief
- Approved sample scripts
- MLR rules and weighted scoring criteria
02
Influencer submits script
- Hawky maps it to the brief, the samples, and the MLR criteria
- AI produces a criterion-level score, a rationale, and flags
03
Human QA reviews the AI result
- Reviewer validates key flags and applies judgment on sensitive MLR items
04
Script outcome
- Green — approved to produce
- Red or needs changes — influencer receives specific edits and resubmits
- Escalate — sent to Abbott MLR or the brand team for a decision
05
Influencer submits final video
- Hawky checks whether the video matches the approved script
- Visuals, spoken claims, text overlays, disclosures, unsafe depictions and the remaining MLR criteria are assessed
06
AI scoring and flagging
- Weighted score is calculated
- A failed high-priority criterion overrides the overall score and triggers a red flag
07
Human QA review
- Reviewer confirms the final result before it is sent externally
08
Video outcome
- Green — ready for Abbott MLR review and the publishing workflow
- Red — returned with timestamped issues and required edits
- Escalate — uncertain or high-risk content goes to Abbott's MLR team
09
Dashboard and audit trail
- Status by influencer and video: submitted, under review, changes requested, approved, escalated
- Scores, violated criteria, reviewer comments and version history stored against every asset
10
Reporting
- Volume, turnaround time, approval and rework rate, common MLR failures, and escalations

The engine

The hard part already runs.

Weighted criteria scoring, beat-level script structure, video entity extraction and queue processing are live in Hawky today. What we build for Abbott is your rulebook on top of them.

Screens shown on other brand accounts · Abbott data has not been ingested

Step 01 · Brand Setup (Demo)Where your rulebook lives. Guidelines, briefs and approved sample scripts are uploaded into structured folders and embedded as retrievable knowledge. Once they are here, every check inherits them by default — and each brand gets its own setup, so Ensure and PediaSure never share a criteria set unless you want them to.

Hawky script view showing beats with spoken lines and on-screen text alongside a weighted criteria score — Steps 03–05 · Script check (Demo)**Weighted criteria scoring, already running.** The right-hand panel is exactly the mechanism your MLR scoring needs: criteria scored individually, the weakest called out by name, and a specific written fix against it. Today these criteria measure performance. **Your medical and regulatory criteria replace them and the machinery is unchanged.** On the left, each beat separates what is **said** from what appears **on screen** from what is **shown** — the three surfaces a claim can hide in.

Hawky Creative Details showing a video creative with extracted entity data — Steps 07–09 · Video check (Demo)**Video read down to its entities.** Running on video, not stills. Speaker type, model gender and age range, language, music or voice, hook type, video format, brand mention timing and CTA timing — who is making the claim, in what language, and at what point in the video. That last part is what makes a flag actionable instead of just a verdict.

Hawky processing 113 assets in one pass with per-row status tracking — At volume (Demo)**Queue processing with per-asset status.** A hundred and thirteen assets moved through in a single pass, each tracked independently to completion. At 120 videos a month this is the difference between a review service and a review bottleneck — nothing sits unassigned and nothing is silently skipped.

Pricing

One rate per video, both checkpoints included.

Priced per video and billed within the platform subscription, rather than as a separate review service.

India · opening market special offer

$14.5per video

Volume: 150 videos a month, per brand
Minimum billing: $2,175 a month, per brand

A special rate for India as our opening market with Abbott. Script check and video check, AI and human review on both.

US & global markets

$16.5per video

Volume: 150 videos a month, per brand
Minimum billing: $2,475 a month, per brand

Applies to the US and all other global markets. Same scope, same checkpoints, same turnaround.

What the rate covers

Both checkpoints, end to end

Brand Setup — encoding your MLR criteria, weights and thresholds, and the brief and sample scripts.
Script check — AI scoring plus human verification, with the failed criteria named.
Video check — script fidelity and visual compliance, flags timestamped.
The approval cycle — runs inside Hawky. Your reviewer sees each asset with its score, flags and evidence attached, and approves or sends it back in the platform. Human in the loop, on Abbott's side.
The evidence trail — a criterion-by-criterion record per asset, handed to your MLR team with the green flag.
Re-scoring — re-submitted content is checked again at no additional charge.

No separate setup fee, no per-seat charge, no charge for red-flag returns.

Scaling

Second brand and beyond

Each additional brand gets its own Brand Setup at the rate for its market, since criteria and claim sets differ by product.
150 a month per brand is the minimum billing volume, and it is sized to carry the 400-in-a-quarter ramp without a re-quote.
Volume above 150 is charged at the same per-video rate — no tiers, no step change.
Escalation volume falls as the criteria calibrate against your reviewers' decisions, so turnaround improves after the first quarter.

Reviewed at the end of the first quarter against actual volume and escalation rates.

Scoring

Criterion-level scores with a written rationale, not a single number.

Override

A failed high-priority criterion triggers red regardless of the total.

Flags

Timestamped to the frame and the line that triggered them.

Fidelity

Final video checked against the approved script, not just against MLR.

Escalation

A defined third path — uncertain content routes to Abbott, never guessed.

Approval

Your reviewer approves inside Hawky, with evidence attached.

Audit

Scores, violated criteria, comments and version history per asset.

Reporting

Volume, turnaround, rework rate, common failures, escalations.

Turnaround

What TAT looks like at 400 to 450.

Turnaround is set by reviewer capacity, and reviewer capacity is something we staff to the ramp. The AI pass does not change with volume; the human QA pass is what we scale.

StageOver a year · ~38 a monthWithin a quarter · ~150 a month

AI pass

Under 30 minutes from submission, script or video

Under 30 minutes — unaffected by volume

Script verdict

Same business day

Within 24 hours

Video verdict

Within 24 hours

Resubmission

Same TAT — re-entering at the AI pass

Escalations

Sits with Abbott MLR — outside our TAT

End to end

From script submission to an approved video, assuming normal influencer turnaround between stages, five to seven business days per influencer — with scripts and videos running in parallel across the cohort rather than in sequence. At the quarter pace, 450 influencers clear in roughly twelve weeks provided submissions arrive evenly; a cohort that all submits in the final fortnight will queue, and that is a scheduling question rather than a capacity one.

Committed TAT is agreed after the calibration batch, once escalation rates are known rather than estimated

Next steps

Four things, and we can start calibrating.

01
Abbott shares the scoring framework
The ten medical and ten regulatory criteria with their weights, the definition of a high-priority criterion, and the brief with approved sample scripts.
02
Hawky encodes it into Brand Setup
Criteria, weights and thresholds configured for the first brand, with your reviewer alongside us so the encoding is confirmed rather than assumed.
03
Calibration batch
A set of scripts and videos your team has already passed and failed. We run ours next to yours — agreement confirms the scoring, disagreement shows exactly which weights need adjusting.
04
Agree TAT and go live
Committed turnaround and escalation thresholds set against the calibration results, then the first cohort runs.