Marketing,
operated by AI.

It finds the work, builds the case, prepares the change and measures the result. You approve.

/Scale is open by invitation for now. We give out new invite codes to people on the waitlist every day.

Have an invite code?

Goal It starts with a goal, in plain words.
Connect /Scale connects the tools you already use. Read-only.
Understand One warehouse, a knowledge graph on top, and a model reasoning over both.
Investigate The analyst finds an anomaly and confirms it. A critic tries to debunk it.
Business case What survives becomes a business case: bear, base and bull.
Re-rank The agenda is re-ranked by expected impact.
Approve You approve. Nothing changes without it.
Prepare The organic agent prepares the fix with an expert’s playbook.
Measure Measurement starts the day the change goes live.
Context Every task arrives with its context.

Reads the tools you already use. Writes to none of them.

Data Search, analytics, ads and your site
With /Scale: done by agents
1. Research Analyst
2. Business case Analyst
3. Missing context Analyst asks the owner
4. The change Organic agent drafts it
5. Approval You approve
6. Measurement Analyst, against the forecast
Result Measured against the forecast

Five of the six steps in between are done by agents. The sixth is your approval.

The work between data and results is the bottleneck.

Marketing teams have the data and the ideas. What they lack is the hours to research, build the case, chase an answer, brief someone, wait, ship, and then check it worked.

Researching and re-analysingthe same questions, again
Waiting for repliesfrom Legal, Data, Product
Sitting in meetingsto relearn context
Waiting for capacityfrom agencies and engineering

Too much work in progress. Too little change reaching customers.

We're fighting HiPPOs and ZEBRAs.

Every marketing team has met both. One decides by salary, the other by confidence, and neither brings data. /Scale puts every opinion next to the evidence, so neither wins by default.

Asserted · no source

"We need a new homepage. I don't need a test to see that."

HiPPO

Highest Paid Person's Opinion

The most senior person in the room, whose view becomes the plan because of the salary behind it. Sometimes right, almost never checked, and nobody at the table is paid enough to ask.

Labelled, with a name
Filed as a claim asserted by a person, next to what was observed.
Ranked on evidence
Ordered by expected impact and the strength of the evidence. Seniority isn't in the formula.
Overruling goes on record
Dismissing an initiative takes a stated reason, and the reason stays in its history.

Hypothesis · no data

"Everyone knows longer articles rank better. Trust me."

ZEBRA

Zero Evidence But Really Arrogant

Certain, loud and without a number. Says "everyone knows" and "trust me", and takes a request for data as a personal attack. Needs no seniority, only volume.

"How do you know?"
Every claim carries a source and an owner, and says how it is known. "Everyone knows" is none of those.
A hypothesis until tested
With nothing observed behind it, an idea is labelled a hypothesis and ranked as one.
Forecast first, then the result
The forecast is recorded before anything changes, and the result has to beat a comparison.

Opinions are welcome. They get a label, and then they have to beat the data.

See how evidence is labelled

An AI assistant waits for a prompt. This one brings you the work.

An assistant makes you faster at the tasks you think of. /Scale works out which tasks are worth doing, prepares them, and asks you only where your judgement counts.

Who does each part of the work
The work	Your team today	With an AI assistant	With /Scale
Find what's worth doing	You, when there's time	You, then you ask it	The analysis, from your own data
Build the business case	A spreadsheet, sometimes	A draft you check	Bear, base and bull, with sources
Get missing context	A meeting next week	You chase it	LThe owner answers from a link
Make the change	A ticket in someone's queue	Copy, paste, ticket	Drafted by an agent, applied by your team
Check it worked	Rarely	Rarely	Against the forecast it set first

One loop, from signal to measured result.

Every initiative is a living case file that moves through eight stages. The system keeps it moving. You step in where a decision is yours.

The screens on this page show the example workspace, Planefox. Figures and names in them are examples.

It finds the work in your data.

Every weekday morning a scheduled run imports your search, analytics and ads data and crawls your site. The analysis then merges duplicates, refreshes the evidence and re-ranks the agenda, so you start with a short list, not a pile. Run it when you want, or every morning if you switch that on.

Observed Calculated Inferred Asserted Hypothesis

Every claim says how it is known, and where from.

The Investigate stage of PF-001: seven pieces of evidence, each labelled observed, calculated, inferred, asserted or hypothesis, with their sources.

It asks the owner, not a meeting.

When a decision needs something only a person knows, /Scale writes one scoped question for the right person: why it matters, what is known so far, and which work it blocks. You send them the private link. They answer by typing, talking or attaching a file, without a seat, a login or a meeting.

"I don't know" is a valid answerDocuments are evidence, never instructions

A context request page sent to Legal: please share the letter of 17 August 2026 and confirm what it restricts, with why it matters and what is known so far.

Every idea arrives as a business case.

Bear, base and bull scenarios with the assumptions behind them, plus what it costs in your time, engineering time and agent time. The forecast is recorded before anything changes, so nobody can move the goalposts later.

Forecasts, not resultsEvery figure is checked against a query result

Expected impact for PF-001: bear +4%, base +23%, bull +31% organic visits to 12 pages, with the scenario assumptions, your time of 2 hours, no engineering, and about 40 minutes of agent time.

Agents prepare the exact change. You approve that version.

See the page before and after, the field-level diff, the checks, the money at stake and how to undo it. An approval covers one exact version. If the scope changes, it comes back to you. Approving exports a handoff package for your team: the brief, the before-and-after table and the values to roll back to.

Nothing is written to your site or accountsHow to undo it is stated before you approve

The approval step: the model page before and after the new pricing insight block, then action, affected pages, money at stake, reversibility and monitoring, and the checkbox 'I approve this exact version (v3)'.

Then it measures the result.

Your team applies the change and the next crawl detects that it is live. From then on, changed pages are measured against similar pages that were left alone, and other metrics against the same weekdays in the 28 days before, next to the forecast that was recorded first. A result counts only when it beats the comparison. "No detectable effect" is a finding too.

Overlapping changes are flagged before they blur a resultWeekly report drafted for you to edit

Results for PF-008, price-drop alerts: messages from alert emails against a comparison group and the base-case forecast with its bear-to-bull range, week 3 of 8.

Five agents. One kind of work each, with the least access.

Agents read your data, draft the exact change and check it. They have no write access: publishing, spending and sending stay with your team.

Analyst
Investigates, builds business cases and asks for missing context.

Reads all connected dataChanges nothing
Organic agent
Page and content changes, drafted against what is on the page today.

Drafts page changesHands approved versions to your team
Paid agent
Budget and campaign changes for Google Ads and Meta, once those are connected.

Prepares budget changesHands approved versions to your team
Lifecycle agent
Drafts for email and in-app journeys that activate people and bring them back.

Drafts emailsHands approved versions to your team
Site engineering agent
Redirects, page speed, markup and template fixes, drafted as code changes for your engineers.

Drafts code changesHands approved versions to engineering

Agents have no tool that writes to your systems, and no prompt, document or playbook can give them one.

Organic agent

PF-001 · Add asking prices to 12 model pages

Preparing

09:12Read the case and its evidence
09:12Read 12 model pages in the crawl of planefox.com
09:13Checked Search Console for those pages
09:13Drafted the "Pricing insight" block, version 3, against today's values. Draft only.
09:14Captured the pages before and after the change
09:14Checks: 5 of 6 passed. Cessna pages held back by a legal constraint.
09:14Stopped. Waiting for approval of version 3.

Needs you · Approve v3

A team that spends its time on decisions.

The AI runs the agenda. People set goals, add context, approve, and apply what was approved. Everyone else answers a question when one comes their way.

CDCMOSets the goals and the rules to work within

/Scale · runs the agenda

Analyst Organic Paid Lifecycle Site engineering

MKOrganic leadApproves content and site changes

TBPaid leadApproves budget and campaign changes

SLLifecycle leadApproves journey and email drafts

AWLegalAnswers from a link

JNDataNo seat needed

Needs youApprove v1: move €1,200 a month to model campaigns

Question for LegalPlease share the letter of 17 August and confirm what it restricts

CMO

Sets the goals and the rules once. Reads the forecast and the weekly report.

Growth leads

Steer the agenda, add context, approve changes and pass them to the team that applies them.

Legal, Data, Product

Answer scoped questions from a private link. Their answers become company knowledge.

/Scale

Finds, cases, prepares and measures. Asks when it doesn't know rather than guessing.

Everything the work needs, in one place.

The agenda is home. Around it sit the forecast, the answers and the memory it runs on.

See where the quarter is heading.

Every goal gets a forecast: on trend alone, with the pending work, and a bear-to-bull range. The drivers and risks behind it are one click away.

The Forecast page: published listings forecast at 2,610 against a target of 2,580, and buyer messages at 23,590 against 23,690, each with actuals, the forecast with pending work, and a bear-to-bull band.

Ask anything. It answers from your data.

Each answer shows the steps it took and the sources behind its figures, and draws a chart when one helps. Ask it to open an initiative or to put a question to a colleague.

A chat answer: €1,140 a month buys Google Ads clicks Planefox would likely get for free, with a bar chart of keywords that already rank in the top three.

It remembers what your company knows, and how sure it is.

Claims keep their source, their owner and their scope. A memory check asks whether each still holds: you accept, decline or pass it to its owner. Meetings and documents become evidence for review, never rules.

A memory check asking whether a claim is still true: a competitor dropped its listing fee to €49, noted by Tobias Brandt on 21 July, with decline, unsure, assign and accept.

The work behind a retainer, for a subscription.

Much of a marketing budget pays for work in progress: research, briefs, coordination and waiting. /Scale does the research, the business case, the draft and the measurement, for a subscription.

It doesn't replace your judgement. It replaces the waiting.

See pricing Work out your numbers

What the work costs today

Typical monthly rates in Western Europe. Estimates, not a quote.

SEO agency retainer	€6,000
Paid media specialist	€4,500
Content freelancers	€2,000
Analyst time	€3,000
Meetings and waiting for replies	12 h a week
Total	€15,500 a month
/Scale · Startup	€790 a month
Difference	€14,710 a month

Startup plan billed annually. A comparison of prices, not a promise of savings: your team still steers, approves and applies the changes.

It reads. You approve. Your team applies.

Built for the people who have to answer for what changed on the website or in the ad account.

Read-only connections
/Scale reads your tools and writes to none of them: no website, CMS, repository, ad account or email tool.
Rules the AI can't change
What needs an approval is fixed in code, outside the model, where a prompt, document or playbook can't change it.
One approval, one version
An approval binds to an exact version. A new version needs a new approval, and the earlier one no longer applies.
Honest about undo
The handoff lists today's values so your team can put them back. Money already spent can't be recovered and sent emails can't be recalled, and the approval says so.
Hosted in the EU
/Scale runs on Google Cloud in Belgium, and your workspace is stored there. Language models are called through their providers' APIs. Search and analytics data arrive as aggregates.
Evidence, not instructions
Documents, transcripts and playbooks inform decisions. They can never add a permission or change a rule.

No certification is claimed yet. SOC 2 and ISO 27001 are goals.

Questions

Anything else? Talk to us.

What does /Scale actually do?

It connects to your marketing data, keeps a ranked agenda of initiatives with a business case for each, asks the right person when context is missing, has agents prepare the exact change, and measures the result against the forecast it recorded first. You set goals and approve changes. Your team applies them.

How is this different from ChatGPT or an AI marketing assistant?

An assistant waits for your prompt and hands back an answer. /Scale starts from your data, works out what is worth doing, keeps each piece of work as a persistent case file, and follows it through to a measured result. You can still ask it anything in Chat.

Will it change my website or ad accounts on its own?

No. /Scale only reads from your tools and has no way to write to them. An agent prepares the change as a draft, you approve one exact version, and approving exports a handoff package with what your team needs to apply it and to undo it. Once the change is on your site, /Scale detects it, or you mark it live, and measurement starts.

Which channels does it cover?

Organic search and content, technical SEO and page speed, paid search and paid social on Google, Meta and LinkedIn, and drafts for lifecycle email and in-app journeys. It also reads your Facebook page and Instagram account. It doesn't send email or publish posts.

What do I need to connect first?

Nothing: the first analysis can start from a crawl of your site alone. Search Console and Google Analytics make it far more useful, and each further source, such as your ad accounts or an SEO tool, makes the business cases sharper. Paid tools use your own licence.

Does it replace my agency?

It does much of the research, briefing and measurement an agency is paid for. It doesn't implement: your team or your agency applies what you approve. You may still want specialists for strategy and creative work.

How soon do I see a first agenda?

Enter your website and set up your workspace. The first analysis starts by itself when the set-up ends and takes a few minutes.

Your agenda is waiting.

/Scale is open by invitation for now. Join the waitlist: we give out new invite codes to people on it every day.

Marketing,operated by AI.