ENFR
Book 30 minutes
Menu
ENFR

A new kind of company has become possible. Make sure the first ones in your market are yours.

Blaast builds and operates complete new businesses beside your core: one orchestration of autonomous agents, coordinated by an AI CEO, outside your systems, under your committee’s mandate.

Book 30 minutes with the founding team Write to us

By video, under NDA if you wish.

Figure 1. A whole business, operated by one orchestration, under your committee’s mandate.

Illustrative example: an architectural model built from bricks. In the centre, on a street, the new business: a lit, glazed ground floor under a blank black sign, and above it four levels of offset white modules, cantilevered. Figures stand at the entrance; a red delivery van waits at the kerb. On the right, a yellow tower crane sets the last module; at its foot, the site yard with its hoarding and site board. On the left, set back, the group headquarters, a grey tower.
  1. 1Your new business, in your name
  2. 2The AI CEO plans and prioritises
  3. 3Your group

Illustrative example

At a glance

  • Most companies have fitted autonomous agents into organisations built without them, and the results have not followed (Figure 2).
  • Blaast starts beside your core: complete new businesses, built and operated by one orchestration and coordinated by an AI CEO, with nothing to integrate (Figures 1 and 3).
  • Each venture is run as an option: a budget capped in the contract, stop criteria written before launch, a decision every month (Figure 5).
  • The code enforces limits, the decision stays with your committee, and everything built is yours (Figures 6 and 8).

Agents have been fitted into old organisations. The results have not followed.

Adoption is no longer the question. McKinsey found about 90 in 100 vertical use cases still stuck in pilot mode, and named the obstacles: poor integration with enterprise systems, fragmented data, siloed teams, organisational inertia [1].

In McKinsey’s 2026 survey, 37 in 100 respondents attribute at least some effect on EBIT to the technology, about the same share as a year earlier [2]. In France, 24 in 100 proofs of concept run with start-ups reach deployment, against 31 in the United States [3].

The usual answer is to integrate harder. The other answer is to start where there is nothing to integrate.

“Nearly eight in ten companies report using gen AI – yet just as many report no significant bottom-line impact.”

McKinsey & Company, QuantumBlack, 13 June 2025 [1]

Figure 2. Use is widespread. Conversion into results is rare.

Use is widespread. Conversion into results is rare.
Share, %Share, %Source
Vertical use cases that “remain stuck in pilot mode”about 90McKinsey, 13 June 2025 [1]
Respondents who “attribute at least some EBIT impact to AI use”37McKinsey, 25 August 2026, 1,719 respondents [2]
Proofs of concept with start-ups converted into deployment, France24Bpifrance Le Hub, 16 September 2026, 437 respondents [3]
Same measure, United States31Bpifrance Le Hub [3]
Scope differs by source: worldwide surveys of companies for McKinsey; collaborations between large companies and start-ups for Bpifrance Le Hub, not businesses created in-house. Declarative surveys.

Start beside the core, not inside it.

Inside the core, the path is long. BCG X found that companies spend “up to three years considering whether to add a new business”, then typically “another three years to commercially launch” it [4].

A business designed around the orchestration from its first day has nothing to integrate, no legacy data to migrate and no organisation to redesign. It needs no connection to your information systems, and the contract says so.

Separate does not mean detached. O’Reilly and Tushman described the companies that manage both: distinct units, “tightly integrated at the senior executive level” [5]. Here, that senior level is your committee.

“Now, AI is shifting the build playing field: lowering experimentation costs, quickening build cycles, and letting AI-native businesses be designed from the outset.”

P. Jenkins and J. Grieve, McKinsey partners, Fortune, 21 September 2026 [6]

Figure 3. Beside the core, there is nothing to integrate, migrate or reorganise.

Bird’s-eye view of the model in the first month. On the left, the headquarters on its dark base; its networks, pale lines, stop dead at the edge of the base. On the right, on its own plot, the glazed ground floor of the new business, the crane and the row of modules, the site hoarding and its board facing the street. A strip of black plinth separates the two bases: no piece links them.
  1. 1Your group: its systems, its data, its teams
  2. 2No connection to your information systems
  3. 3Commissioned by your committee

Illustrative example

Two ways to put autonomous agents to work
Two ways to put autonomous agents to workInside the coreBeside the core
SystemsIntegration with existing enterprise systemsNone: no connection to your information systems, written into the contract
DataLegacy data, fragmented across functionsData created by the venture, and yours
TeamsSiloed teams to coordinateOne orchestration, coordinated by an AI CEO
DecisionProject reporting, milestone by milestoneEvidence from the market, decided every month by your committee
Path“Up to three years” to decide, “another three years” to launchA real business in its market, with a date for the decision
Obstacles as named by McKinsey [1] and BCG X [4]; project reporting as described by McKinsey partners [6].

Figure 4. Work flows from one task board, and the plan changes when results come in.

One orchestration runs the whole business.

The orchestration does not answer requests one at a time. It builds the business, from the offer to the measurement, then keeps operating it.

Step 1: on the plot, behind the hoarding, the glazed ground floor of the new business is laid piece by piece, then its sign and the first level of modules, outlined as new pieces.
  1. 1The functions, laid one by one
The functions, one by one: the shopfront and the sign, then the modules of the first level. Each piece is set in its place.

It operates continuously.

Work flows from a shared task board and is dispatched automatically. The AI CEO re-plans when results come in.

It covers every function of a young company.

Strategy, finance, product, front-end and back-end engineering, infrastructure and deployment, security, marketing, search, content, translation, sales, analytics, and quality assurance on prospect and customer journeys.

It remembers.

Memory persists from one work cycle to the next: the venture’s business context, its task board and the full history of changes.

Step 2: the yellow crane rises in the yard, section by section; at its foot, four modules wait in a row in the storage area. New pieces outlined.
  1. 1One task board
The AI CEO: the crane serves the row of modules, the single task board all the work starts from.

It closes on results.

A task tied to a business objective cannot be closed without a measured result attached. Otherwise it stays open, and only a person can cancel it.

Step 3: figures arrive at the entrance, and the crane lifts the next module. New pieces outlined.
  1. 1Demand, measured at the door
The results: demand arrives at the entrance, and the crane lifts the next module. The plan changes when results come in.

Illustrative example

Every venture is an option: a capped premium, a date, a decision.

When building a business costs less, the sound response is not to spend less. It is to open more options, and to commit only to those the market confirms.

Each Blaast venture is written as an option, in four terms of the contract (Figure 5).

“[…] lower cost per experiment is not an argument for cutting venture budgets. Instead, it is an argument for running more experiments.”

McKinsey & Company, 31 March 2026 [7]

Figure 5. The premium is capped, the date is set, and each month the decision is your committee’s.

Premium
The budget, capped in the contract. Each venture is priced on its own scope.
Underlying
A real business in the real market, under your brand or a neutral brand. You validate the offer and the brand before it goes live.
Expiry
Stop criteria and a date, written before launch.
Exercise
At each Monthly review, your committee takes one of four decisions.

At each Monthly review, your committee takes one of four decisions. Nothing is glued, so each one remains possible.

Close-up of the plot on Review day: the crane has set a third level of modules on the building, outlined.

Continue

Same mandate, next month. One more level is set, outlined.

Close-up of the plot: the second level of modules is set in another position, cantilevered to the left, outlined.

Adjust

The offer, price or target changes. Here the second level is set differently, outlined.

Close-up of the plot: the building stays standing, closed, a grey shutter down over the shopfront, outlined; the crane and the row of modules are gone.

Stop

A stop report. The building stays standing, closed, the crane gone: what was built stays yours.

Close-up of the plot: a second crane goes up at the back, and two more levels, outlined, crown the building.

Accelerate

More budget, or handover to your teams: a second crane goes up and two more levels are set, outlined.

Contract terms shown as a structure. Each mandate sets its own date. Illustrative example.

An AI CEO coordinates. Your committee governs.

The AI CEO is the first management role of this new kind of company. Its written mandate is to plan and prioritise; it excludes writing code, deploying, sending email and touching payments. In its reports, cost and activity figures are computed from the operating database, not written by the model.

A person can steer or pause a venture at any time. And every month, at the Monthly review written into the contract, your committee decides what happens next (Figure 6).

Figure 6. The decision is your committee’s. The contract and the code set the limits.

Who holds what, for each venture

Your committee decides

  • The mandate: hypothesis, market, brand, budget cap.
  • Every month, at the Monthly review: continue, adjust, stop or accelerate.
  • When to hand the business over to your teams.

The contract commits

  • A budget capped in the contract.
  • Stop criteria and a date, written before launch.
  • Offer and brand validated by you before going live.
  • Your brand or a neutral brand, under confidentiality.
  • No connection to your information systems.
  • Code, content, accounts and data are yours.

The code enforces

  • A cost ceiling and a time limit on every work cycle; a cycle that exceeds either is stopped and its partial work recorded.
  • Daily activity limits per function and a global concurrency limit, whatever the model decides.
  • One gateway per venture for deployments, DNS changes, outgoing email, payment-catalogue changes and secret reads; each call is written to an audit log.
  • Each venture kept apart: its own access token, secrets, deployment space and domain, browser profile and encrypted credential vault.
  • An append-only operating journal: the database rejects any modification or deletion of a recorded event.
  • Pausing a venture stops all its activity and closes its gateway access.

The AI CEO coordinates, within its written mandate

  • Plans and prioritises; re-plans when results come in.
  • No code, no deployment, no email, no payments.
  • Cost and activity figures in its reports are computed from the operating database.

Contract terms are engagement commitments. Code mechanisms are capabilities of the technology.

Businesses you could open beside your core.

Figure 7. Each venture is judged on the strongest evidence of demand it can produce.

Two examples for each leadership team. Each shows a structure: the market movement, the business you open, what your committee will know and when. None shows a result.

General management

The business you open
Win back shoppers lost to low-cost platformsA fashion retailer sells a short range of basics online, under a neutral brand, shipped quickly from France.Illustrative example
Market movement
Since 1 July 2026, the EU charges a flat €3 customs duty per item category on imported parcels worth €150 or less [8].
What your committee will know, and when
Whether new customers pay, then buy again, by day 60, against criteria written before launch.
Expected signal
Payment
The business you open
Sell in Germany before opening a subsidiaryA family-owned biscuit maker opens a German-language store with local payment methods and capped campaign spend, shipped from its stock in France.Illustrative example
Market movement
55% of French mid-sized companies say weak demand weighs on their business [9].
What your committee will know, and when
Paid orders, what each cost to acquire against the margin, and reseller requests, at the date set in the contract.
Expected signal
Payment Qualified request

Innovation leadership

The business you open
Fixed-price home repairs, without an insurance contractAn insurer sells a pay-per-call emergency repair service under a neutral brand, delivered by its own network of tradespeople.Illustrative example
Market movement
56% of executives report a low appetite for reputational risk, up from 36% in the previous survey [10].
What your committee will know, and when
Paid call-outs, and the share of customers who agree to hear about an insurance offer, by day 60.
Expected signal
Payment Sign-up
The business you open
Three offers in parallel, one committeeA contract catering group opens three businesses built on its central kitchens at once: meals delivered to older people, buffets for small companies, a student meal plan.Illustrative example
Market movement
59% of leaders whose companies launched three or more ventures in five years draw more than 10% of revenue from them, against 32% where only one was launched [11].
What your committee will know, and when
Which of the three produces paid, repeat orders, compared on the same criteria at each Monthly review, where budget moves to the strongest.
Expected signal
Payment

E-commerce leadership

The business you open
Which second-hand model, for which customers?A baby-goods retailer compares three formats online: trade-in for store credit, refurbished items, and the sale of unsold stock.Illustrative example
Market movement
41.4% of French online shoppers bought at least one second-hand product over twelve months [12].
What your committee will know, and when
Which format produces paid sales. Each has its own stop criterion, written before launch.
Expected signal
Payment Qualified request
The business you open
Sell where your customers now ask their questionsA garden retailer sells project kits on pages written for conversational assistants, beside its main site and without touching its systems.Illustrative example
Market movement
Since late August 2026, ChatGPT has offered advertising in France [13].
What your committee will know, and when
The sales truly added, measured against a control group, at the date set in the contract.
Expected signal
Payment
Fictitious companies. Orders and requests reach your teams through a channel agreed at framing and written into the contract, with no connection to your information systems. Market facts are public, dated and sourced; thresholds are set at framing.

What stays yours. What we don’t do.

Yours, written into the contract

Code
Yours.
Content
Yours.
Accounts
Domain, payment, advertising: in your name.
Data
Yours. Nothing is taken from your systems.
Brand
Yours, or a neutral brand that never exposes it.
Confidentiality
A confidential contract, and an NDA before the first meeting if you wish.
Exit
Planned from the first day.

What we don’t do

  • We don’t bring the market to you. Our autonomous agents build and operate. The audience comes from your market access and a campaign budget you approve.
  • We don’t connect to your information systems. You share an offer through your own channels. Orders and requests reach your teams through a channel agreed at framing.
  • We don’t take on high-risk uses. No hiring, credit or health ventures, as European regulation defines them.
  • We don’t publish prices. Each venture is priced on its own scope, with a budget capped in the contract.

Figure 8. At exit, the AI CEO comes off. Everything else stays in your name.

Handing over the keys, at the end of the day: the new business is finished and open, its shopfront lit under the sign, figures at the entrance, the van in the yard. The hoarding is gone. On the side street, on the right, the folded crane leaves on its convoy.
  1. 1Stays in your name: the business, its code, its data
  2. 2The AI CEO steps away at exit

Illustrative example

In 30 minutes, see how an operated business is built and run.

By video, with the founding team.

We show you how a venture is framed, built and operated, with an illustrative mandate and Monthly review.

We look together at where a first venture could sit beside your core. Under NDA if you wish.

For the briefing

  • 1×growth question you want to examine
  • 1×decision-maker on your side
  • 30minutes, by video

Write to us

Prefer to write first? Tell us what you would like to discuss.

Or by email: hello@blaast.app

Sources

  1. McKinsey & Company, QuantumBlack, “Seizing the agentic AI advantage”, 13 June 2025.
  2. McKinsey & Company, “The state of AI in 2026: On the road to ROI”, 25 August 2026. 1,719 respondents, fieldwork from 4 May to 8 June 2026.
  3. Bpifrance Le Hub, “Collaborations startups - grands groupes : la fin de la POC Culture ?”, 16 September 2026. 437 respondents (227 start-ups and scale-ups, 210 large companies) in France, Germany, the United States and Singapore; declarative answers.
  4. BCG X, “How Companies Can Speed Up the Business of Business Building”, 15 November 2023.
  5. C. O’Reilly and M. Tushman, “The Ambidextrous Organization”, Harvard Business Review, April 2004.
  6. P. Jenkins and J. Grieve, McKinsey partners, op-ed in Fortune, 21 September 2026.
  7. McKinsey & Company, “How to build businesses faster and better with AI”, 31 March 2026.
  8. French Ministry of the Economy and Finance, press release no. 850, 30 June 2026.
  9. Bpifrance Le Lab, 16th barometer of French mid-sized companies (ETI), 23 June 2026. 534 responses.
  10. WTW, 2026 Reputational Risk Readiness Survey, 28 April 2026. 500 executives responsible for risk strategy in 20 countries, companies with revenue of at least 1 billion US dollars.
  11. McKinsey & Company, “The way to win in corporate venturing: Serial building and AI”, 28 October 2025. 715 respondents from companies with revenue above 100 million US dollars.
  12. Fevad, key figures 2026, reported by E-Commerce Nation, 1 October 2026.
  13. Blog du Modérateur, 17 August 2026, updated 19 August 2026.