AI AgentHypotheses & variants

Discover Amir

CRO cycle · 02–03 · Cadrage, brief & design

Amir turns frictions into hypotheses backed by past tests, prioritizes them into a roadmap and breaks them down into variants ready to test. Each hypothesis displays its evidence.

Recalls tests already run on the same lever
Rejects unsupported hypotheses
Prioritizes your backlog into a roadmap
Helps design the test: KPI, duration, pitfalls

Book a demo See the whole fleet

Amir

Hypothèses & variantes

Driven by a Welyft CRO expert

Real-world use cases

What Amir can do for you

Scenario 1 · Hypotheses for a page
Amir
Live
You : Suggests 5 prioritized hypotheses for this cart page, with evidence from the database.
09:41 · Seen
AmirHere are 5 hypotheses in the if… then… because… format, ranked by priority. Each one displays its evidence from the database: concluded tests, winners, robust, and a solid, indicative, signal, or absent level.
Scenario 2 · Challenged backlog
Amir
Live
You : Challenges this backlog of 12 hypotheses and flags tests already run.
11:08 · Seen
AmirI sorted each hypothesis: already validated, mixed, contradicted, or new. Duplicates are flagged, and hypotheses contradicted by the database are removed from the roadmap with the reason.
Scenario 3 · Updated roadmap
Amir
Live
You : Updates the quarter's roadmap with the results of the last 5 tests.
14:26 · Seen
AmirThe roadmap incorporates the 5 results: confirmed levers move up, contradicted leads move down, and tests already run are flagged to avoid relaunching them.

Illustrative examples: each mission starts from your data and your tools.

Deployment

Agents trained like experts, updated like products.

Each agent is a versioned plugin on GitHub. Three ways to make the most of it.

During a mission

During the engagement

Our consultants call on the agents at every stage of the CRO cycle. You receive the deliverables, without installing anything.

Via the Welyft platform

Remote, at Welyft

The fleet runs on our platform, connected to your tools with your consent. An expert validates every deliverable.

Internalized within your organization

ConstructionWelyft builds the agent using its method, test learnings and CRO best practices.

PersonalizationEach agent is tailored to your method, brand guidelines and vocabulary.

GitHub MarketplaceYour custom agents install as plugins from a repository dedicated to your organization.

UpdatesEvery Welyft agent update is versioned, reviewed and delivered to you.

Your toolsConnected to your stack via MCP.

GuardrailsA fixed permissions matrix that decides and validates.

Within your organization

We build your agents using our method, customize them, then a GitHub plugin marketplace delivers their updates.

Let's talk about your environment Train your teams: AI for CRO

Watch Amir turn friction points into prioritized hypotheses.

Amir, Welyft's hypotheses & variants agent, grounds every hypothesis in our past tests, prioritizes it into the roadmap, and breaks it down into variants ready to test.

Book a demo

Its skills

Support, prioritize, break down

Skills · Context

Live
Past tests recalled
Amir relies on a RAG of over 270 anonymized A/B test results and a bank of 300 UX recommendations and user test results.
/amir:context
Live
Mei's friction points as input
Mei's friction map becomes the starting point for hypotheses, combined with your analytics data.
/amir:context
Live
Your constraints integrated
Amir connects to your roadmap tool and to the repository of your already launched A/B tests to challenge the roadmap.
/amir:context

Skills · Validation

Live
Supported hypotheses
Amir applies CRO best practices and shares them with your teams, starting with hypothesis building: if… then… because…, with a visible level of evidence.
/amir:validation
Live
Challenged backlog
Already validated, mixed, contradicted or new: each idea is sorted, and duplicates are flagged.
/amir:validation
Live
Test design
Main KPI, duration, variants and pitfalls: at least 2 weeks of testing, in complete weekly cycles.
/amir:validation

Skills · Production

Live
Prioritized roadmap
Hypotheses are ranked according to their evidence and expected impact, flagging tests already conducted.
/amir:production
In progress
Copy variants
Each selected hypothesis is broken down into variants linked to the KPI they target. The dedicated feature is coming soon.
/amir:production
In progress
Mockups and tickets
Clickable mockup of the variant and development tickets: these building blocks are under construction.
/amir:production

Human first

Amir prepares, the Welyft expert decides.

Deliverable: Backlog priorisé et variantes
The expert decides: quoi tester, dans quel ordre, et la variante qui part

Its safeguards

Zero invention: every figure cited comes from the database.
Reasons by KPI level (micro, intermediate, macro), against the baseline rate of the same level.
Each variant is linked to the hypothesis and the KPI it targets.
On the client side, only aggregated patterns from at least 3 tests are shared.

Under the hood

What Amir relies on

Trained on

Nos méthodes de priorisation et de statistique, et les best practices métiers

Data used

Welyft A/B learnings base (around 270 tests)
Welyft CRO audits and UX best practices
Mei's friction map
Client analytics data
Client backlog, roadmap, style guide and constraints

Specialties

/amir:welyft-ab-patterns
Variantes · coming soon
Copy · coming soon

It's live : /amir:welyft-ab-patterns

In progress : Copy variants, clickable mockups and automatic creation of development tickets are still to be built.

Our offers

Offers where Amir is involved

FAQ

Frequently asked questions about Amir

Une question sur votre contexte ? Nos experts vous répondent.

Posez votre question

Where does Amir's evidence come from?

From the Welyft A/B learnings database: around 270 anonymized tests, with their uplift rate by lever and by KPI level. Amir does not cite any figure that doesn't come from it.

Are other clients' tests exposed?

Amir reasons on anonymized learnings. On the client side, only aggregated patterns from at least 3 tests are shared, with no test identifier.

Can we hand over our existing backlog to Amir?

Yes. Amir sorts each hypothesis (validated, mitigated, contradicted or new), flags duplicates and tests already run, then proposes a roadmap.

Who chooses what gets tested?

The Welyft CRO expert decides what to test, in what order and which variant goes live. Amir prepares the options and their evidence.

Do agents have access to our data?

Only to the sources you open, with a scope and access defined by contract. Agents work on your tools and your exports; no personal data is copied into a deliverable.

In a few seconds

Amir, from data to deliverable

A backlog of hypotheses, challenged

Hypotheses & variants

Prioritized with our A/B learnings

Backlog priorisé et variantes

Priorisé avec nos learnings A/B.

A prioritized backlog

3 hypotheses

Reassurance below the CTAP1
Delivery from the product listP2
Multi-step formP3

~270

Welyft tests in the learnings database

uses: audit, challenge, patterns, test

Win rate by lever

What we've already tested, lever by lever

Page audit: URL, screenshot or mockup

Backlog prioritization

Anonymized database, no client mentioned

Illustrative demonstration: the examples are not client results.