Architecture, templates and most of the content are already produced

This is the single most important fact in this plan. Jesse and Chanti have produced the menu architecture, the page templates and the bulk of the content; the Keyrus AI manifesto copy has been live on the dev instance since 6 July (43 entries, 508 fields, 7 languages). The 18 September gate is therefore a validation, not a production phase, which is what makes the December release reachable.

Preprod, updated 26 October: keyrus-ai.netlify.app, built in Next.js and connected to the keyrus-ai Contentful environment. It replaces the former dev instance, keyrus-redesign.netlify.app on development-3. Status of each page: Status 26 Oct.

From validation on 18 September to release on 3 December

Proposed sequence, to be adjusted and signed off. Validation steps, then development, then the full delivery track: alpha, UAT, a beta phase with its own UAT layer, and a production rollout. Development is 15 working days, Monday 21 September to Friday 9 October. Scroll the chart horizontally on narrow screens.

Exec validation cadence

Run by Stephan and Christophe, roughly every two weeks. The Shift workshop in Paris is the one in-person gate, with all Shift leaders.

Validation process, in order

Click a step for its owner and validation loop.

Four people join the setup in September

On group marketing, on the content that is not industry specific, and on customer references. Industry experts are mobilised per industry once the template is validated.

Joining

In place

Who produces what, with which dependencies

The critical path runs through the 18 September validation: it releases development and the remaining content production. Most workstreams are already well advanced, so what follows is mostly finalisation and sign-off rather than production from scratch.

6 decisions still open, 5 already settled

Choices and comments are saved live in this browser, and the summary copies in one click for the write-up. Recommendations are proposals from the project team.

Fourteen more, outside the numbered decisions

Flagged on the ClickUp sitemap rather than in the decision list. They are structural and editorial calls that the mockups cannot answer on their own. The menu shared on 8 September already settles one of them.

• Structure
  Two sitemap blocks flagged Can these be consolidated?
• Structure
  Resources to collapse into a single page. The menu shared on 8 September already does this: Insights carries Videos, Events and Customer Stories & Use Cases.
• Structure
  Some sections will not be pages of their own.
• Contentful
  One page or several to find content? Asked with an explicit doubt about what the CMS can do. It governs the four filters.
• Partners
  Group partners by HOM layer instead of alphabetically? The industry pages already do it, with an Engines row under each capability.
• Editorial
  The three matrices still have to be produced. Nothing renders properly without them.
• Languages
  How much gets translated: the whole site, or the core?
• Design governance
  Who controls the thumbnails and headers that will surface in the homepage cards? Raised on 18 September by Jesse and Chanti. The mitigation on the table is centralised templates plus closer work with group marketing. Until it exists, the cards should carry text only.
• Localisation
  The homepage cards are localised per country. Who publishes into them, with what editorial rule, and what happens in a country with no event that month? Christophe: it should carry insights, news and campaigns, not events only, or the block goes stale.
• Process
  Focused ownership instead of full attendance. Raised by Baptiste on 18 September: with a fixed deadline, an executive room where everyone arrives with their own approach slows every round. One focus per person, Brahim on portfolio, Catherine on use cases, with this app as the shared level of view. To settle with Christophe and Stephane before the Paris workshop.
• Menu
  A visible client entry? The three-entry baseline puts use cases and customer stories inside Insights. On the table: a curated Clients entry that opens the use case index filtered on customer stories.
• Labels
  Videos names the format, not the message. Demos or Media proposed.
• Homepage
  Show use cases before capabilities in the homepage order?
• Portfolio
  Cybersecurity by design to be made explicit in the Trust & Governance capability. Brahim, with the portfolio.

Delivery model: the Loop replaces Advise · Implement · Operate

Stage names to confirm.

• Decision
  The Loop replaces the three-step delivery model: Advise, Build, Embed, Prove (names from the portfolio of 19 October; Diagnose, Activate, Orchestrate, Prove on 8 October), carried by one team with five roles and no handoff. Stage names to confirm. Source: SHIFT pre-read, Pillar 1 v2, a proposal for discussion.
• Nearshoring
  Open. The mockups show nearshoring as a transverse element, with no page and no menu entry. Whether it gets a dedicated page is to settle. Nearshoring covers who staffs the Loop team, adapted by country.
• Services
  The menu label stays Services, to confirm. The page for each stage is to design: no screen exists yet.
• Screens
  The Services captures (services-0, menu3-services and the board thumbnail) predate 8 October and show the old model. They are not regenerated.

Five reusable templates, now spread over twenty screens

Produced by Jesse and Chanti, exported on 18 September: the thirteen mockups of 8 September plus the seven editorial templates, and a reworked homepage. Offer copy is written and final, editorial copy is still lorem by design. What follows is the reading, the defects and the calls it leaves open.

The above the fold, settled on 18 September

The one design point that was still open going into the executive review. Christophe had raised it twice: the hero is beautiful, and nothing on it makes you want to scroll beyond the wow effect.

The four structural shifts

Fix before it goes outside

Six placeholder slips carried by the mockups. None of them is a design problem; all six are copy that was never written or was pasted twice. Re-checked against the 18 September export: the About hero still reads “Talk to our insurance team”. The other five have not been re-verified screen by screen and should be before development starts.

• Industry
  AI & agentic systems repeats the Data foundations bullets, word for word. Its own copy was never written. The rewrite is drafted in Capabilities by page v1.
• About
  The hero button still reads Talk to our insurance team.
• About
  The Collective intelligence card repeats the hero standfirst.
• About
  All three HOM layers carry the same description. The industry page gets it right.
• Partners
  Axon Technologies is captioned twice, once under the Dataiku logo.
• Figures
  Two formats for the same numbers: 28 / 2.8k / €331M against 30+ / 2,800+ / €331.7M.

Pre-plug the default, then give the experts one week to correct it

Two documents dated 26 August cover all sixteen pages: Outcome levers by page v2 for section 02, Capabilities by page v1 for section 06. Between them, 80 outcome rows, 380 action levers and 256 capability points, all counted against their character limits. Nothing here is a placeholder.

Browse the drafted content

Open a page to read its five outcomes with their value lever line and action levers, then the four capability cards with their points and engines. This is the copy the experts are asked to correct.

Who owns which page

The mapping that has to exist before the window opens. Names are typed here and kept in this browser; the summary copies out for the chat.

Verification queue

Twenty-three items the drafts flag themselves, most of them regulatory names or partner engines. Legal review applies to every regulation named, and partner tier wording is unconfirmed everywhere. S02 is the outcome levers element, S06 the capability cards.

Twelve calls that apply to every page

Raised by the two documents themselves. They are resolved once and applied across the set, so they belong in the 18 September session rather than with the individual experts.

Three top-level entries, the baseline from the executive review

The 18 September export had six entries. The executive review cut it to three: What we do, Insights, Who we are. What we do opens on Industries and now holds Technologies and Services. Use cases and resources merge into Insights. HOM, the group pages and Careers sit under Who we are. Chanti redrew it on 24 September, and it is confirmed at the Paris workshop.

01 · What is inside it

The menu rebuilt, so you can open it. One panel at a time behaves like the real thing. Everything open gives the whole architecture on one screen, which is the view the menu itself never provides.

02 · Why it is arranged this way

Each panel carries its own justification. What follows is what the executive review closed, and what it leaves open.

The panels, as designed

Redrawn by Chanti on 24 September, on the three-entry baseline. The live menu above carries the same content.

The closed bar

What we do · Industries, default panel

What we do · Business functions

What we do · Capabilities

What we do · Technologies

What we do · Services

Capture before 8 October: old services model. The live menu above shows the four Loop stages.

Insights

Who we are

HOM becomes the backbone of the site

The corporate presentation V1 (June 2026) sets the narrative: Architect of intelligence, carried by the Human Orchestrated Model™. The site must reflect this architecture: HOM's 3 layers, the portfolio of 4 capabilities, and the taglines sprinkled across all pages. It's also the GEO lever: a proprietary, structured narrative is exactly what AI engines reproduce best.

The narrative's taglines, to highlight and sprinkle across the whole site (hero, sections, CTAs, offer pages).

Open thinking: making HOM demonstrable

The concern raised by the exec team: the positioning is a little too intellectual. HOM and “architect of intelligence” land at manifesto level, and the business directions need real examples to see themselves in it. Not decided, and not blocking the build.

1,087 existing pieces of content, already mapped to HOM

The full portfolio audit (Jesse, June 2026): 553 blog insights, 434 key plays, 100 EPM pages. The 677 thematic content pieces spread across HOM's 3 layers in a balanced way (39 / 30 / 31 %): HOM isn't a narrative to fill in, it's a grid that structures editorial capital that's already there. The remaining 410 (136 industries, 85 events, 69 corporate, 19 client cases, 101 out of scope) cross-tag to a pillar or fall outside scope.

This mapping changes the approach: HOM isn't landing on an empty site. Every surface (HOM, portfolio, industries, insights) can be linked from V1 to existing content tagged by layer, and the editorial calendar can be driven by the grid's gaps rather than ad hoc. That's decision D9.