Skip to content
Mariete

Every product. Every real-world case.

49 runbooks across the Mariete suite. Real inputs, real process, real outcomes — straight from the field, indexed by product and function.

§ The suite

Eleven products.
Pick one to read.

Click a tile to narrow the page to just that product’s runbooks. The two flagship products lead; the rest follow in release order. Click again to show everything.

Chapter 01
5 cases
Builder.
Does the tasks.
Builder · 5 runbooks

An autonomous task agent. Here’s what it ships.

Every runbook below has three phases — the inputs you provide, the work Builder actually does, and the output a real team uses. Sourced from customers running this today.

Run Builderbuilder.mariete.ai
Sales ops·Featured runbook
Builder
A SaaS sales team

Automated lead research and enrichment before stand-up.

Every Monday the team feeds Builder a list of 500 target company names. Builder visits each website, pulls LinkedIn, identifies the right contact, enriches with firmographic data, scores against the ideal-buyer profile, and writes a personalised opener — all before morning stand-up.

Outcome ↑
8 min
vs 3 hours manual
OutreachEnrichmentCRM sync
Open runbook
01Input
Drop 500 target companies in (CSV or HubSpot)
02Process
Lead-research agent scores and enriches each one
03Output
Personalised opener lands in the SDR’s queue
Competitive intelBuilder
08:00
Delivery time every Monday

Weekly competitive monitoring, delivered by 8 AM Monday.

A strategy team

Builder’s research agent runs every Monday morning. It scrapes competitor pricing pages, job boards, press releases and review sites published in the prior week, then compiles a structured 2-page briefing and emails it to the team — no manual aggregation, no missed updates.

01
Input
Watch list: competitor domains + review sites
02
Process
Research agent scrapes and summarises the week
03
Output
2-page briefing in the team inbox, 08:00 Monday
BriefingsWeeklyAutomation
Open
OutboundBuilder
3×
Reply rate at same headcount

Full-cycle outbound sequences that triple reply rate.

An e-commerce brand

Builder’s outreach agent runs end-to-end cold email: it segments the list, writes personalised openers referencing each prospect’s recent news, schedules sends at peak times, auto-follows up non-openers on day 3 and 7, and escalates warm replies to a human rep. Reply rate triples; SDR capacity holds flat.

01
Input
Import the list, connect mailbox + CRM
02
Process
Agent drafts, sends and follows up per persona
03
Output
Warm replies escalate to a human closer
Cold emailSequencesEscalation
Open
Content opsBuilder
5×
Channels from one post

Turn one blog post into five channels of content.

A content marketer

Publish one long-form post per week. Builder’s content-distribution agent automatically repurposes it into a LinkedIn article, three X threads, a short-form newsletter and a Slack community post — each rewritten for the platform’s tone and scheduled at peak engagement times.

01
Input
Publish one long-form post
02
Process
Agent rewrites for 5 channels in-voice
03
Output
Five pieces scheduled at channel-specific peak times
RepurposeMulti-channelScheduling
Open
SupportBuilder
3 min
First response, down from 4 hr

Tier-1 support that resolves 68% of tickets itself.

A 60-person software company

Builder’s support agent reads incoming tickets, classifies by issue type, pulls relevant help docs, and either resolves automatically or routes to the right teammate with a pre-drafted reply. Average first response drops from four hours to three minutes.

01
Input
Inbound ticket hits the shared queue
02
Process
Agent classifies, pulls docs, drafts reply
03
Output
68% auto-resolve — the rest route, pre-drafted
TriageZendeskAuto-resolve
Open
Chapter 02
5 cases
Inspector.
Does the thinking.
Inspector · 5 runbooks

A deep-research agent that reads, reasons and cites. Here’s what it ships.

Every runbook below has three phases — the inputs you provide, the work Inspector actually does, and the output a real team uses. Sourced from customers running this today.

Run Inspectorinspector.mariete.ai
Fundraise·Featured runbook
Inspector
A founder preparing for Series A

Pre-fundraise health check, two months before pitching.

Inspector runs parallel deep-dive assessments across eight core departments and ten strategic areas. It flags that unit economics are solid but the churn methodology is non-standard — the founder fixes the gap before any investor meeting, reducing diligence friction.

Outcome ↑
18
Dimensions scored
FundraiseDiligenceMetrics
Open runbook
01Input
Describe the company, stage and goals
02Process
Eight departmental agents run in parallel
03Output
Evidence-backed health report with priority flags
Board prepInspector
90 min
From full-day board prep

Monthly executive narrative generated from raw KPIs.

An operations lead

Paste in the month’s numbers. Inspector produces a plain-language summary: what improved, what declined, what the leading indicators suggest, and what decisions need to be made now. Board prep drops from a full day to 90 minutes.

01
Input
Paste raw KPIs at month-end
02
Process
Inspector structures the narrative and flags
03
Output
Board deck draft, cited and sourced
Board deckKPIsNarrative
Open
OperationsInspector
$340K
Recoverable annual margin

Surface the $340K leak in your delivery process.

A 120-person services firm

Inspector’s operations and finance agents analyse project data, utilisation rates and billing patterns. They surface three specific workflow breakpoints and estimate $340K in recoverable annual margin — enough to rebuild the ops team around.

01
Input
Connect project, staffing and billing data
02
Process
Ops + finance agents analyse utilisation
03
Output
Three breakpoints + recoverable-margin estimate
OperationsMarginServices
Open
M&AInspector
20 min
Per target, vs 2 weeks

Screen 18 acquisition targets in 20 minutes each.

A PE-backed roll-up

For each target, the deal team inputs publicly available data and Inspector produces a standardised 8-dimension health score. It surfaces the top four targets for deeper diligence — replacing two weeks of analyst work per target with a 20-minute process.

01
Input
Upload public data per target
02
Process
Inspector scores 8 dimensions
03
Output
Top-4 ranked shortlist for deep diligence
Private equityScreeningM&A
Open
Market readinessInspector
4
Readiness dimensions scored

Catch that you’re product-ready but sales-unready.

A software company entering a new vertical

Inspector evaluates positioning, sales capability, pricing compatibility and support infrastructure. The assessment reveals the team is product-ready but sales-unready — saving them from a premature, expensive push into the vertical.

01
Input
Describe the target vertical
02
Process
Readiness agents assess four dimensions
03
Output
Structured go / no-go with gaps listed
GTMReadinessRisk
Open
Chapter 03
5 cases
Painter.
AI marketing asset builder.
Painter · 5 runbooks

A visual node-flow workspace. Here’s what it ships.

Every runbook below has three phases — the inputs you provide, the work Painter actually does, and the output a real team uses. Sourced from customers running this today.

Run Painterpainter.mariete.ai
Launch·Featured runbook
Painter
An e-commerce brand launching a sneaker line

Product-launch campaign asset set in one session.

Painter produces the full campaign in one session: 6 Instagram feed posts, 3 Stories templates, 2 LinkedIn banners, 1 email header, and 4 ad creative variants — all in the brand palette, ready to schedule. Three days of freelance-designer work done in 40 minutes.

Outcome ↑
40 min
Full campaign, vs 3 days
LaunchCampaignMulti-format
Open runbook
01Input
Lock brand kit (logo, colours, fonts, tagline)
02Process
Wire inputs → generation → outputs on the canvas
03Output
16 on-brand assets ready to schedule
Agency opsPainter
4 hr
Turnaround, vs 48 hr

Eight client brands, one workspace, zero bleed-through.

A digital agency managing 8 client brands

Painter runs with a separate brand kit per client. When a brief lands, the team selects the client’s kit, drops the brief in, and generates six variants in the correct visual identity. Client approval rates climb and turnaround drops from 48 hours to 4.

01
Input
Onboard each client brand kit once
02
Process
Brief → matching pipeline generates variants
03
Output
On-brand outputs, 4-hour turnaround
AgencyMulti-brandBrand kits
Open
SocialPainter
3×
Posting frequency, same headcount

Always-on social calendar with a 15-minute approval.

A SaaS marketing team

Every Monday the team inputs the week’s theme and three key messages. The pipeline auto-generates 5 LinkedIn posts, 7 X posts, and 3 short-form video scripts, all consistent with brand voice. The CMO reviews and approves in 15 minutes. Posting frequency triples without adding headcount.

01
Input
Drop in theme + 3 messages on Monday
02
Process
Pipeline generates 15 assets in brand voice
03
Output
CMO approves in 15 min; posting tripled
SocialAlways-onBrand voice
Open
CreativePainter
3
References → originals

Turn top competitor campaigns into on-brand originals.

A marketing team refreshing creative

Painter’s discover surface pulls real competitor campaigns by industry and platform. Before each campaign, the team browses what’s performing, selects three references, and Painter generates original assets inspired by the aesthetic — lifting measurable creative quality.

01
Input
Browse discover by industry + platform
02
Process
Select 3 reference campaigns as mood
03
Output
Painter generates originals in the aesthetic
CreativeCompetitiveMood
Open
SeasonalPainter
960
Assets across 3 campaigns

Reskin 80 templates for Black Friday in two hours.

A retailer with 12 product categories

Painter’s batch mode runs 80 existing templates through a "holiday reskin" node. 960 assets across Black Friday, Christmas and New Year are generated in two hours — no rebuilding from scratch.

01
Input
Point batch mode at existing template set
02
Process
Apply the "holiday reskin" node
03
Output
960 assets across 3 campaigns generated
SeasonalRetailBatch
Open
Chapter 04
5 cases
Forecast.
Visual time-series forecasting.
Forecast · 5 runbooks

Drag nodes, wire pipelines, run predictions. Here’s what it ships.

Every runbook below has three phases — the inputs you provide, the work Forecast actually does, and the output a real team uses. Sourced from customers running this today.

Run Forecastforecast.mariete.ai
Supply chain·Featured runbook
Forecast
An online home-goods brand

SKU-level demand forecasts, with holidays as covariates.

Forecast predicts demand for each of 240 SKUs for the next 12 weeks with confidence intervals. The ops team right-sizes purchase orders, reducing overstock by 31% and stockouts by 44% in Q4.

Outcome ↑
240
SKUs forecasted, −31% overstock
RetailInventoryCovariates
Open runbook
01Input
Load 18 months of SKU sales into a node
02Process
Attach holiday + promo calendar as covariate
03Output
12-week forecast per SKU, confidence bands
Anomaly detectionForecast
1.5σ
Deviation alert threshold

Catch churn anomalies 3–4 weeks earlier.

A SaaS company watching churn

Forecast learns the normal churn pattern and alerts customer success whenever actual churn deviates more than 1.5 standard deviations from baseline. The team intervenes with at-risk accounts three to four weeks earlier than before.

01
Input
Pipe weekly churn and MRR into Forecast
02
Process
Anomaly node learns the baseline
03
Output
Alert fires at >1.5σ deviation
ChurnAnomalyEarly warning
Open
FinanceForecast
6 mo
Rolling horizon, auto-refresh

Rolling 6-month revenue forecast, rebuilt monthly.

A 75-person services firm’s CFO

Forecast pulls invoicing data from accounting software via API, cleans it for seasonality, and produces a 6-month rolling revenue forecast by service line. The pipeline runs on the 1st of each month — board meetings no longer open with a 30-minute argument over which Excel is correct.

01
Input
Pull invoicing data via accounting API
02
Process
Clean for seasonality; forecast by service line
03
Output
6-month rolling forecast, auto-refreshed
CFOCash flowBoard
Open
CapacityForecast
£18K
Monthly saving across 6 sites

Staffing rotas that save £18K a month across six sites.

A restaurant group with 6 locations

Forecast predicts footfall by location and daypart for the next four weeks using historical covers, reservations, and a local-events calendar. Managers set rotas with 25% less overtime while maintaining service — saving roughly £18K a month across the group.

01
Input
Connect covers, reservations, events calendar
02
Process
Forecast by location + daypart, 4 weeks out
03
Output
Managers set rotas; overtime −25%
HospitalityStaffingRotas
Open
MarketingForecast
84%
Historical scenario accuracy

Model budget shifts before you commit the spend.

A head of growth

Forecast the scenario: "what if we shift 20% of paid search budget to LinkedIn?" Feed in 24 months of weekly spend and lead data across six channels; the model predicts MQL impact with 84% historical accuracy. Allocation decisions are data-backed in hours, not weeks.

01
Input
Load 24 mo of channel spend + leads
02
Process
Run scenario on the modelling node
03
Output
MQL impact predicted at 84% accuracy
GrowthAllocationScenarios
Open
Chapter 05
2 cases
Cascade.
Map the consequences.
Cascade · 2 runbooks

Six specialist agents stream a navigable consequence tree for any event. Here’s what it ships.

Every runbook below has three phases — the inputs you provide, the work Cascade actually does, and the output a real team uses. Sourced from customers running this today.

Run Cascadecascade.mariete.ai
StrategyCascade
3rd
Order depth mapped

Map second- and third-order consequences before you ship.

A leadership team reviewing a pricing change

Six specialist agents stream a navigable consequence tree for any event — a pricing change, a rollout, an acquisition — so the exec team walks in already knowing what breaks three moves out.

01
Input
Describe the event in plain English
02
Process
Six agents fan out consequence branches
03
Output
Navigable tree, depth-3 by default
StrategyPricingRisk
Open
PolicyCascade
120+
Ripple nodes surfaced

What-if engine for the policy team.

A policy lead drafting a regulatory change

Draft a regulatory change in plain English. Cascade traces ripple effects across stakeholders, timelines and knock-on markets — with every claim cited back to a source document.

01
Input
Paste the draft proposal
02
Process
Cascade runs cited stakeholder branches
03
Output
120+ ripple nodes with citations
PolicyStakeholdersCited
Open
Chapter 06
2 cases
Nexus.
A living knowledge graph.
Nexus · 2 runbooks

Ingest notes and vaults. Here’s what it ships.

Every runbook below has three phases — the inputs you provide, the work Nexus actually does, and the output a real team uses. Sourced from customers running this today.

Run Nexusnexus.mariete.ai
KnowledgeNexus
5 yr
Of tribal knowledge indexed

Turn five years of notes into one searchable graph.

An ops lead on a 40-person team

Point Nexus at the shared drive. It extracts people, projects, decisions and dates into a temporal graph, then answers questions on top — every answer grounded in the source document.

01
Input
Point Nexus at the drive
02
Process
Extract entities into temporal graph
03
Output
Ask questions; answers cite their source
KnowledgeGraphCited
Open
LegalNexus
T−30
Days before every trigger

Contract graph that pings before every renewal.

An operations + legal lead

Point Nexus at the contracts folder. It builds a living graph of parties, clauses, obligations and renewal dates — and pings Slack 30 days before every trigger.

01
Input
Connect the contracts folder
02
Process
Nexus extracts parties + obligations
03
Output
Slack ping, 30 days before every trigger
ContractsRenewalsObligations
Open
Chapter 07
5 cases
Atlas.
Market-entry intelligence.
Atlas · 5 runbooks

Describe your business, pick a target market. Here’s what it ships.

Every runbook below has three phases — the inputs you provide, the work Atlas actually does, and the output a real team uses. Sourced from customers running this today.

Run Atlasatlas.mariete.ai
Expansion·Featured runbook
Atlas
A UK HR-software company

Feasibility study on three markets in one afternoon.

Instead of three consulting engagements, the team runs Atlas deep-analysis on Germany, Netherlands and Spain in parallel. Within four hours they have scored reports on regulation, competitive density and recommended local partner types. Germany scores 81/100 and becomes market one. Total cost: $297, vs a comparable $75K spend.

Outcome ↑
$297
vs $75K consulting comparable
ExpansionRegulationGTM
Open runbook
01Input
Describe the business + pick target markets
02Process
Eight specialist agents analyse in parallel
03Output
Scored reports with citations, in 4 hours
GTMAtlas
1
GTM pivot avoided the wrong channel

Catch the WhatsApp-first Brazil GTM before you spend on ads.

A US SaaS company entering Brazil

Atlas’s consumer-intelligence agent reveals Brazilian B2B buyers rely heavily on warm referrals and prefer WhatsApp-based sales communication over email. The company redesigns its Brazilian GTM entirely — partner-led sales, WhatsApp integration — before spending on ads.

01
Input
Describe product + target market
02
Process
Consumer-intelligence agent returns channel map
03
Output
GTM redesign decision, pre-spend
GTMLATAMChannels
Open
RegulatoryAtlas
8
Markets screened in one run

Screen 8 APAC markets for data-localisation friction.

A US HealthTech startup

Atlas quick-scans 8 APAC markets filtered for data-localisation laws and healthcare AI regulation. It surfaces Singapore and Australia as lowest-friction entry points, and flags Japan’s 18+ month PMDA approval path — saving months of misdirected legal spend.

01
Input
Pick 8 APAC markets to screen
02
Process
Regulatory agent filters for data laws
03
Output
Ranked friction scores + timing flags
RegulatoryAPACHealthTech
Open
PE diligenceAtlas
4
Markets re-scored in diligence

Catch regulatory instability in an acquisition target’s market.

A PE firm evaluating a logistics SaaS

Atlas runs comprehensive analyses on the target’s four operating markets. It surfaces one market with regulatory instability that makes the business model legally fragile, and another where three well-capitalised local competitors recently entered — the deal team re-prices the valuation.

01
Input
Run Atlas on the target’s 4 operating markets
02
Process
Eight agents score each dimension
03
Output
Regulatory + competitive flags → valuation update
Private equityDiligenceMulti-market
Open
White spaceAtlas
3
White-space markets identified

Find strong-demand / weak-supply markets you’d missed.

A fintech founder scoping expansion

Atlas scans 12 emerging markets for a specific use case: SME lending infrastructure. The competitive-landscape agent identifies Egypt, Vietnam and Nigeria as strong-demand / weak-supply. The founder adjusts the roadmap to target Vietnam first, backed by Atlas’s consumer intelligence on local financial behaviour.

01
Input
Scan 12 emerging markets by specific use case
02
Process
Landscape agent maps demand vs supply
03
Output
Roadmap prioritised on the gap
FintechWhite spaceEmerging
Open
Chapter 08
5 cases
Radar.
Competitive intelligence, automated.
Radar · 5 runbooks

Add competitors by URL. Here’s what it ships.

Every runbook below has three phases — the inputs you provide, the work Radar actually does, and the output a real team uses. Sourced from customers running this today.

Run Radarradar.mariete.ai
Pricing·Featured runbook
Radar
A SaaS marketing team

Catch a silent 25% competitor price cut before sending the newsletter.

Minutes before the monthly newsletter goes out, Radar fires an alert: the primary competitor just quietly dropped their Pro tier price by 25%. The team updates the comparison page and reframes messaging before send — a potential embarrassment turned into a timely response.

Outcome ↑
25%
Competitor cut — caught pre-send
PricingReal-timeAlerts
Open runbook
01Input
Add competitors by URL
02Process
Collectors monitor pricing + announcements
03Output
Alert fires → team reframes the send
Real-timeRadar
3
Enterprise trials that week

Turn a competitor outage into three enterprise trials.

A marketing team watching competitors

Radar surfaces a competitor outage discussed on Reddit and X in the morning briefing, with sentiment analysis and an estimated affected-user count. Marketing publishes a targeted "reliability-first" ad within hours — three enterprise trials start that week.

01
Input
Morning briefing lands at 08:00
02
Process
Outage detected with sentiment + user count
03
Output
"Reliability-first" ad live in hours
Real-timeSentimentResponse
Open
StrategyRadar
6 mo
Lead time vs public news

Read competitor strategy six months before the press release.

A VP of Strategy

Radar monitors job postings across eight competitors. When one starts posting aggressively for "ML Infrastructure" engineers, Radar flags it as a strategic signal — six months before any public announcement. The team adjusts the roadmap and partnership strategy accordingly.

01
Input
Watch list: 8 competitors’ job boards
02
Process
Signal detection flags role-type surges
03
Output
Roadmap + partnership updates pre-announcement
SignalsHiringStrategy
Open
M&ARadar
3 mo
Lead time on the announcement

Spot market consolidation three months before the deal.

A CEO monitoring the competitive set

Radar monitors 15 competitors for M&A signals: drops in job postings, funding silence, founder "new chapter" LinkedIn posts, unusual patent assignments. When two mid-tier competitors show three signals simultaneously, the CEO proactively engages investors — three months before the acquisition announcement.

01
Input
Watch 15 competitors across 4+ signal types
02
Process
Radar correlates drops, silences and moves
03
Output
Early warning → investor conversations
M&ASignalsConsolidation
Open
Sales enablementRadar
+18%
Win rate, one quarter

Live battlecards in 30 seconds per competitive deal.

A 90-person B2B sales team

Two hours of manual battlecard prep per deal turns into 30 seconds. Radar auto-generates feature comparison, pricing, known weaknesses from reviews, recent news and talk tracks. Win rate against the top two competitors improves 18% in one quarter.

01
Input
Competitor named on a deal
02
Process
Battlecard auto-generates from live signals
03
Output
Rep walks in with talk tracks + weaknesses
BattlecardsSalesWin rate
Open
Chapter 09
5 cases
Pulse.
Feel your audience before launch.
Pulse · 5 runbooks

Synthetic audiences of AI personas react to your content in 60 seconds. Here’s what it ships.

Every runbook below has three phases — the inputs you provide, the work Pulse actually does, and the output a real team uses. Sourced from customers running this today.

Run Pulsepulse.mariete.ai
Creative·Featured runbook
Pulse
A DTC brand

Don’t split $40K across two creatives — pick the winner in 4 minutes.

Instead of splitting budget and waiting two weeks for data, the team runs both creative routes through Pulse against 120 synthetic personas matching their target. Pulse returns a clear winner in four minutes. Full budget goes to the winning creative — 2.3× ROAS vs historical average.

Outcome ↑
2.3×
ROAS vs historical
CreativeMediaROAS
Open runbook
01Input
Submit both creatives
02Process
Pulse tests against 120 personas
03Output
Clear winner in 4 min → full spend
SegmentsPulse
+41%
Developer-segment demo bookings

Make one landing page work for two very different buyers.

A B2B SaaS company

Pulse runs the same landing page against "operations manager" personas and against "CTO" personas. The ops segment converts; the CTO segment is confused by technical vagueness. The team ships a CTO-specific variant — demo bookings from the developer segment jump 41%.

01
Input
Submit one page, two persona sets
02
Process
Pulse returns segment-specific friction
03
Output
Ship a segment variant where it matters
B2BSegmentsLanding
Open
LifecyclePulse
+34%
Average open rate, 6 months

Subject-line testing that climbs open rate 34% in six months.

An email marketer

Eight subject-line variants run through Pulse every send, tested against a 50-persona audience — returning open-rate predictions, emotional tone scores, and spam-risk flags. Over six months, average open rates climb 34%.

01
Input
Generate 8 subject-line variants
02
Process
Pulse scores open-rate, tone, spam risk
03
Output
Ship the top-scoring line; iterate
EmailCRMLifecycle
Open
NamingPulse
3
Names pressure-tested pre-launch

Catch the hidden "complexity" connotation before launch.

A product team deciding a feature name

Three name options run through Pulse against 80 end-user personas. Pulse returns memorability scores, "what do you think this does?" accuracy ratings, and emotional associations. One option scores high on memorability but carries unexpected "complexity" connotations — caught before launch.

01
Input
Submit 3 name options
02
Process
Pulse scores memorability + associations
03
Output
Name chosen on data, not on vibes
NamingProductLaunch
Open
PitchPulse
87
Traction-slide conviction score

Slide-by-slide investor feedback before any real meeting.

A startup founder

Each section of the pitch deck runs through Pulse with investor personas configured. Pulse returns slide-by-slide feedback: which slides build conviction, which generate scepticism, which confuse. Traction scores "87 — strong conviction"; market-size scores "52 — unclear methodology". The founder rewrites before any real meeting.

01
Input
Submit deck slide-by-slide
02
Process
Investor personas react per section
03
Output
Conviction scores per slide → targeted rewrite
FundraisePitchRehearsal
Open
Chapter 10
5 cases
Simulate.
Multi-agent social simulation.
Simulate · 5 runbooks

Source material in, hundreds of cognitive agents out. Here’s what it ships.

Every runbook below has three phases — the inputs you provide, the work Simulate actually does, and the output a real team uses. Sourced from customers running this today.

Run Simulatesimulate.mariete.ai
Messaging·Featured runbook
Simulate
A fintech launching a product

Test two positioning angles against 400 target-segment agents.

Instead of A/B-ing on live audiences, the team runs both positioning angles — "savings-first" vs "freedom-first" — in Simulate against 400 agents matching their target. Simulate reveals "freedom-first" wins for under-35s but triggers a "sounds risky" faction at 45+. Ad targeting segments accordingly.

Outcome ↑
400
Synthetic agents per test
PositioningSegmentsLaunch
Open runbook
01Input
Submit both positioning angles
02Process
400 target-segment agents react
03Output
Segment-specific targeting decisions
Market entrySimulate
1
Rewrite before any ad spend

War-game your UK narrative before a penny of spend.

A US SaaS company entering the UK

British professional agents consistently respond negatively to superlative-heavy US-style copy ("the most powerful", "industry-leading") and respond better to understated competence framing. The team rewrites the UK landing page before spending on UK acquisition.

01
Input
Submit current messaging for target market
02
Process
Market-matched agents react at scale
03
Output
Landing-page rewrite before ad spend
GTMNarrativeUK
Open
ProductSimulate
3
Strategies tested overnight

Choose the rollout strategy that doesn’t spark a vocal faction.

A consumer app tightening free-tier limits

Before shipping, the PM runs three rollout strategies through Simulate: immediate cut-off, 30-day grace period, and grandfathering. Simulate predicts strategy (a) triggers a vocal negative X faction that influences 12% of new signups to delay. The team ships strategy (c).

01
Input
Define 3 rollout strategies
02
Process
Simulate forecasts faction dynamics
03
Output
Pick the strategy with lowest blast radius
PricingRolloutCohorts
Open
Public affairsSimulate
3
Key influencer types identified

Identify the three influencer types that decide the narrative.

A lobbying firm modelling policy reaction

Simulate models public and media reaction to a proposed autonomous-vehicle regulation. Six hundred agents — politicians, journalists, consumer advocates, general public — run for 90 simulated days. The model identifies the three key influencer types whose early stance most determines the broader narrative.

01
Input
Describe the proposed regulation
02
Process
600 agents simulate 90 days of reaction
03
Output
Three key influencer types named
PolicyPublic affairs90-day sim
Open
CrisisSimulate
Day 20
Brand-recovery break-even

Rehearse three breach-disclosure strategies overnight.

A healthcare company anticipating a disclosure

Three response strategies run through Simulate. Strategy (b) — proactive detailed disclosure with customer compensation — results in net-positive brand recovery by day 20. Strategy (a), the minimal legal statement, festers into sustained negative narrative for 60 simulated days. The team pre-drafts strategy (b) before anything happens.

01
Input
Draft 3 response strategies
02
Process
Simulate runs 60-day agent reactions
03
Output
Pre-draft the winner; stand by
CrisisCommsHealthcare
Open
Chapter 11
5 cases
Edge.
Your mind, extended.
Edge · 5 runbooks

A fully offline AI assistant running on-device. Here’s what it ships.

Every runbook below has three phases — the inputs you provide, the work Edge actually does, and the output a real team uses. Sourced from customers running this today.

Run Edgeedge.mariete.ai
Confidentiality·Featured runbook
Edge
A consultant handling NDA’d client material

Draft confidential client work without cloud risk.

Edge runs entirely on the device — no internet, no sync, no telemetry. Sensitive client documents never leave the laptop. The consultant drafts memos and analysis on-device; encrypted biometric-locked storage keeps everything private.

Outcome ↑
0
Bytes leaving the device
On-devicePrivacyConsulting
Open runbook
01Input
Drop sensitive docs into an Edge Space
02Process
Draft and analyse entirely on-device
03Output
Nothing syncs to a cloud provider — ever
Field workEdge
Offline
Full workflow, no connectivity

An AI assistant that works where the signal doesn’t.

A civil engineer on rural infrastructure projects

Edge writes technical reports, analyses field observations, and drafts client updates on-site with no connectivity. Spaces keep project contexts separate. When connectivity returns, the engineer sends polished reports, not rough notes.

01
Input
Open the project-specific Space offline
02
Process
Draft and analyse in the field
03
Output
Sync polished output when signal returns
OfflineField workSpaces
Open
Air-gapEdge
Weeks
Context retained, air-gapped

Context retention across meetings in air-gapped rooms.

A compliance officer in regulated environments

Edge’s persistent memory builds context across weeks of meetings even without any network connection. Notes from Monday’s compliance review flow into Thursday’s audit prep — all on-device, all encrypted, nothing ever leaving the room.

01
Input
Take notes in-room across meetings
02
Process
Edge retains context across Spaces
03
Output
Cross-reference weeks later, still offline
CompliancePersistentSpaces
Open
KnowledgeEdge
Months
Of input, never sent to cloud

A private second brain that isn’t training anyone’s model.

A senior researcher

The researcher pastes notes from meetings, research papers and book highlights into Edge over time. Persistent memory builds an evolving model of their knowledge base. When drafting a synthesis document, Edge surfaces relevant context from months of input — none of it sitting on a cloud provider’s training dataset.

01
Input
Paste notes, papers, highlights over time
02
Process
Edge builds an evolving on-device memory
03
Output
Draft syntheses with cited local context
ResearchMemoryPrivate
Open
LegalEdge
Privileged
Matters drafted off-cloud

Draft privileged legal research without touching the cloud.

A lawyer working on privileged matters

Edge drafts research memos and analysis on-device for matters under privilege. Encrypted storage + biometric lock meet the confidentiality bar no cloud tool can match — the document is written, reviewed and revised without once touching an external server.

01
Input
Open a privileged-matter Space
02
Process
Draft and refine the memo on-device
03
Output
Export the final doc locally, no sync
LegalPrivilegedOn-device
Open
Case in focus · Radar

One Monday morning on Radar’s competitor desk.

A VP of Revenue at a 90-person SaaS company walks into stand-up with the briefing already open on her phone. Fourteen signals, six sources, one playbook. This is what the last twelve hours looked like inside Radar.

Radar’s collectors and synthesis agents work overnight so the briefing is done before you are. By Monday 08:00 it has already filtered the firehose down to what actually moved.

radar · monday-briefing.logLIVE
22:14collector.news·indexed 312 articles across 10 watched competitorsnews
01:08collector.reddit·pulled 47 threads mentioning competitor-a pricingreddit
03:41synth.pricing·merged 3 sources into one signal: enterprise tier −18%synth
05:22collector.jobs·flagged 4 new EU sales reqs at competitor-bjobs
06:55synth.weekly·drafted 3-paragraph summary + action cardsynth
07:58notify.slack·dropped the briefing into #competitive-desksink
§ Ship with us

Pick the case that looks like your week.

Start with the runbook closest to the work on your desk today. Mariete is one click away.

Get started Book a walkthrough