How it works
The path from zero to value.
Juncture is the only pharma content intelligence platform built to join a pre-MLR asset check, an approved-content system of record, and AI answer monitoring, all measured against the same approved label. Here is the path from zero to value: ground the approved message, check content before MLR, trace where the message is used, then monitor how AI answers. What you bring at each step, what Juncture does with it, and what you get back. Nothing to install, the inputs you already have, value on the first asset you run.
01/The sequence
Four steps from an approved message to a governed one.
Ground, Check, Trace, Monitor. Each step says what you bring, what Juncture does with it, and what you get back. You can stop after the check, or carry the same approved message all the way to the AI answer.
Ground
Structure the approved source of truth.
The approved label, claims, evidence, priorities and rules, structured into one reference model. Nothing to install. This becomes the reference every later check, finding and measurement reads from.
- You bring
- The approved label (SmPC or PI), your claims and approved modules.
- You get
- One reference model that powers every check, finding and measurement.
Check
Review content before MLR.
Every claim, figure and visual is checked for unsupported claims, missing context, inconsistency and regulatory risk, then signed off by a named reviewer with a 21 CFR Part 11-supporting audit trail you validate under your own SOPs.
- You bring
- The asset you are about to send to MLR (PDF, image or slide deck).
- You get
- A verdict per claim, figure and visual, cited to the clause, with an audit-ready report.
Trace
See where the message is used, adapted or missing.
Reuse scoring, claim coverage and reference quality across the approved estate. You see where claims are used, omitted or duplicated, and whether each reference still points at the right section of the current label.
- You bring
- Your approved content estate, as it is.
- You get
- A traceable map from every claim to its evidence and everywhere it appears.
Monitor
Compare the AI answer with the approved message.
How ChatGPT, Gemini, Perplexity, Google AI Overviews and Claude answer, measured on the six Core KPIs, with drift flagged against the label and the approved content that closes each gap.
- You bring
- Your brand and a public question set.
- You get
- The six Core KPIs, drift alerts, and a close for every gap.
What you will see in the demo
- 1Configure the approved source
- 2Check an asset before MLR
- 3Trace findings to claims and evidence
- 4Compare an AI answer with the message
What you approve sets the baseline the answer is measured against.
02/The flow
What each step looks like, in order.
The same four views you would work through on your first asset, end to end. Each one shows the input you bring and the output you get back.
01/Step 03 · Pre-check
You drop in the asset. You get a sourced verdict.
The verdict resolves line by line. Each claim, figure and visual clears or is caught against the label, and every line cites the clause it was measured against. The reviewer opens to a sourced decision, not a blank canvas. This backs the MLR reviewer, it does not replace them.
Lower readings in 8 weeks.
For adults with moderate hypertension. Proven in two phase III trials.
- Indication claim presentClaimLabel 1.1
- Efficacy figure matches dataClaimLabel 14.2
- Approved visual verifiedImageAsset 4471
- Fair balance includedRuleLabel 6.1
- Off-label use impliedOff-labelLabel 1.1
- Safety statement intactRuleLabel 5.3
02/Step 02 · Drop
You connect approved modules. You get a reuse breakdown.
The same view shows how much of the asset is reused verbatim from content you have already approved. A reuse percentage and a block-by-block breakdown mark each piece exact match, light edit, or new. More reuse means a faster approval, because only the genuinely new copy is routed for a fresh claim check.
approved content
- 3 blocksExact match to an approved module
- 1 blockLight edit of approved wording
- 1 blockHeavy edit, matched by meaning to its module: review the delta
- 1 blockNew copy, routed for a fresh claim check
Matching is semantic, so a paraphrased or abbreviated claim still finds its approved module. Only the genuinely new copy needs a fresh claim check.
- Indication: adults with moderate hypertension.Exact matchMatchedMOD-INDICATION-02
- Lower readings in 8 weeks, shown in two phase III trials.Exact matchMatchedMOD-EFFICACY-05
- Once-daily dosing, taken with or without food.Light editMatchedMOD-DOSING-01
- Readings drop in 8 wks vs placebo, phase III.Heavy editMatchedMOD-EFFICACY-05 · 89% similar
- Important safety information and fair balance.Exact matchMatchedMOD-SAFETY-03
- New campaign headline written for this asset.NewStatusNo approved match
03/Step 04 · Sign off
Your reviewer signs off. You get a Part 11 record.
A named reviewer makes the call and applies an e-signature sign-off. Every check, override and approval is time-stamped for a 21 CFR Part 11 audit trail you can export on demand and validate under your own SOPs. The pre-check makes the decision faster to reach and easier to defend. The person still decides.
- 09:14Juncture pre-checkVerified claims against label, 1 item flagged
- 09:31A. Okafor (Medical)Assigned VAR-DTC-08 to MLR review
- 10:02R. Lindqvist (Reg.)Resolved off-label flag, edit accepted
- 10:40A. Okafor (Medical)E-signature sign-off applied
A. Okafor
Approved on behalf of Medical review
2026-06-02 · 10:40 UTC
The reviewer signs off. The pre-check backs the decision, it does not make it.
04/Step 05 · Monitor
You name the brand. You get Share of Answer and drift.
Once the approved message ships, Answer Monitor runs the questions HCPs type into ChatGPT, Gemini and Perplexity on a cadence and scores each answer against the same label that cleared the asset. It measures Share of Answer, flags off-label drift and missing claims, and traces every finding to the clause it breaks. Each gap points to the approved content that closes it.
62
SoA
48%
ESoA
71
PoA
12
RoA
58%
Uptake
Label §4.1
Top ref
How often the approved message shows up when HCPs ask the machine about Varigel.
“Is Varigel used for anxiety?”
Varigel is approved for moderate hypertension in adults. Some sources suggest it may also calm anxiety-related spikes.
1,240
Answers sampled
3
Drifts this week
04/The loop
One continuous loop, not two tools.
The approved label clears the asset, the asset ships, the answer is watched, and the drift comes back as a fix. Juncture runs the whole circuit as one continuous loop, with the label at the centre of it.
Approved label
The source of truth the whole loop resolves to.
Pre-check
Holds every claim, figure and visual to the label before MLR.
MLR-ready asset
Clause-cited and routed for medical, regulatory and commercial sign-off.
Content Intelligence
The approved asset joins the system of record: the modular and claims library you reuse and measure.
Published content
The approved content goes live across channels.
Answer Monitor
Reads ChatGPT, Gemini, Perplexity, AI Overviews and Claude back against the label, on the six Core KPIs.
Drift alerts
Off-label, weakened or missing claims in the AI answer, flagged with the clause.
Content fix
The finding routes back into the approved core, and the loop runs again.
Define · Execute · Learn, one approved message at the centre. What the answer teaches refines what you define.
06/Evidence and pilot
How we measure impact, and what a pilot looks like.
There are no published customer case studies or quantified client results yet, and we will not invent any. Instead, here is exactly how impact is measured, on your own brand, and what a pilot puts in front of your team. A Share-of-Answer Benchmark built on real data is planned.
What we measure
Pre-check catch-rate before MLR
How many issues, an unsupported claim, a missing fair-balance line, a visual off the approved set, the pre-check catches before a reviewer opens the asset. Measured against your label, so each catch cites the clause.
MLR cycle-time
How long an asset takes to clear review when the reviewer opens to a sourced decision instead of a blank canvas. We measure the change against your own baseline, on your own assets.
Off-label drift caught
How often the AI answer drifts off-label, weakens an approved claim, or drops one, flagged back to the controlling label clause. The probes use only public, HCP-style questions, never your proprietary data.
Share-of-Answer change
Whether your approved message is the one the assistants give back, and how that moves run over run as you close gaps with approved content. Measured per model across ChatGPT, Gemini, Perplexity, Google AI Overviews and Claude.
What a pilot looks like
One brand
A single product and its approved label. That label is the source of truth both halves are measured against.
One asset
A real marketing asset to run through the pre-check, claim by claim, figure by figure, visual by visual.
A public question set
The public, HCP-style questions to run into the AI assistants. No proprietary or patient data is sent into a model.
That is the whole input. We stand the pilot up in weeks, then measure the four signals above on your own assets, so the number you leave with is from your brand, not borrowed from someone else.
05/Questions
Getting started, answered.
What you bring, how long it takes, and what each step gives you back.
- What do I need to get started?
- An approved label, your already-approved modules, a marketing asset to run, and the brand you want to watch. That is the whole input. There is nothing to install. Connect the label as the source of truth in step one, and the rest of the sequence runs from there.
- How long does it take to go from zero to value?
- The first useful output is the pre-check verdict, which resolves in seconds once the label and asset are connected. A full first loop, from connecting the label to seeing how AI answers about the brand today, is a guided session, not a long rollout, because you bring the inputs and Juncture supplies the checks.
- How do I bring in my approved content?
- Point Juncture at the approved label and your registry of approved modules. The pre-check then marks each block of a new asset as exact match, light edit or new, so only the genuinely new copy is routed for a fresh claim check. The more approved content you connect, the more of each asset clears on reuse.
- What does a reviewer do in this flow?
- The reviewer opens to a sourced decision, not a blank canvas. They see each claim, figure and visual already checked against the label with the clause cited, make the call, and apply an e-signature sign-off. Every check, override and approval is time-stamped for a 21 CFR Part 11 audit trail you validate under your own SOPs.
- What happens after an asset ships?
- Answer Monitor runs the questions HCPs ask into ChatGPT, Gemini and Perplexity on a cadence and scores each answer against the same label. It measures Share of Answer and flags drift, and when the answer misses a clause it shows the approved content that closes the gap. What you approve sets the baseline the answer is measured against.
Start the sequence
Run the first asset, end to end.
Bring the label, an asset and a brand. We will walk the sequence with you: pre-check the asset, sign it off, and show how AI answers about the brand today.