There is a small temporal problem inside every large company. The company lives in the present. The internet does not. A departed executive remains “current” on an old conference page. A decade-old dispute keeps its excellent search rank. A strategic pivot exists in the annual report but nowhere else. Then an artificial-intelligence model arrives, reads this untidy archive, and speaks with the calm authority of a receptionist who has never once checked the calendar.
This is the problem Vectors has chosen to sell against. The Paris firm calls its latest practice “AI Reputation,” but the name is almost too neat. What it is really describing is the industrialization of memory. A company no longer has only customers, journalists, regulators, staff, and investors to persuade. It also has machines assembling an answer from whatever evidence they can retrieve.
Vectors is built for that messy audience. Its roughly 50-person roster mixes communications consultants with monitoring analysts, designers, videographers, data scientists, developers, and project managers. The menu runs from crisis counsel and public affairs to websites, employee advocacy, executive positioning, social listening, and an in-house recording studio. It looks less like a classic press office than a newsroom, research desk, and production company sharing a very handsome lease near the Centre Pompidou.
A business built in the repair shop
The origin matters. In 2009, Albéric Guigou and Fabrice Ivara founded Reputation Squad, an early French e-reputation agency and the business from which Vectors evolved. Back then, online reputation was often imagined as a repair trade. Something unpleasant appeared in search results; specialists tried to displace, correct, answer, or remove it. The internet was a set of surfaces to manage.
That model met its sharpest limit on Wikipedia. In 2020, French journalists reported that volunteer editors had connected undisclosed paid edits to Reputation Squad and other communications firms. Wikipedia’s machinery is designed to preserve the history of a page, including attempts to prettify it. The medium remembers the repairman as clearly as the repair.
The episode offers a useful lesson without requiring a morality play: secret control is brittle on a public ledger. It also makes the present-day Vectors proposition more interesting. Today’s agency emphasizes measurement, source mapping, crawlability, coherent publishing, and human analysis. Its stated technology rule is restrained: give clients what they need when they need it, “not when the hype dictates.” The work has moved from changing a surface to understanding the system that produces the surface.
The modern reputation problem is not simply what was said. It is what the machine can still find, combine, and repeat.
Thirty-nine thousand eight hundred questions
In the second edition of its CAC 40 AI Reputation Barometer, produced with sister company Trickstr, Vectors asked four models 39,800 questions and analyzed 572,796 citations. Those numbers are not proof that a model “believes” anything. They are evidence of a repeatable communications audit: ask consistent questions, collect the answers, map the sources, compare companies, and watch the result over time.
The crucial finding is mundane enough to be actionable. Models often favor material that answers a question directly, not material that is newest or institutionally official. A beautifully designed corporate site may be almost useless if its claims are hard to crawl or coyly written. A clear third-party page can outweigh this week’s press release. The organization the model describes may be a snapshot from 2020 through 2024, complete with people who left and troubles that never did.
How the second audience forms an opinion
Vectors sells the response in five parts: teach teams how these systems form an answer; diagnose the organization’s portrayal; map the sources behind it; compare the official narrative with the machine’s version; then publish and monitor a broader, multi-source record. Trickstr supplies the large-scale testing platform. An academic partnership with Université Paris Dauphine gives the methodology another layer of scrutiny. Vectors says it already works on the subject with Veolia, Vinci, L’Oréal, and Carrefour.
The unbundled agency, bundled again
AI Reputation is the new door, not the whole house. Vectors’ older advantage is integration. Consider its public account of more than a decade working with Veolia: monitoring, a multilingual and multicultural social ecosystem, content, executive thought leadership, the Verso YouTube channel, websites, and crisis management. Any one of those can be bought from a specialist. The point is to keep them from contradicting one another.
The same logic appears in its Renault Group case. When Renault unveiled a new corporate purpose in 2021, Vectors did not merely write a launch line. It helped recruit and support employee ambassadors across five brands, turning an abstract promise into testimony from people doing the work. The method is easy to copy: find credible internal voices, involve them in construction rather than distribution, train them, and give them material worth sharing.
It also has clear limits. This approach works best for organizations with a substantial public footprint, senior sponsorship, patient access to internal experts, and enough publishing authority to correct gaps across many channels. It works poorly when the underlying conduct remains unchanged, when legal or security constraints prevent useful disclosure, or when leaders expect a single optimized page to overpower a long and well-documented record. A measurement system can reveal incoherence. It cannot make incoherence disappear.
The copyable move
Audit the received story before rewriting the official one.
Ask the same practical questions across search, major AI models, journalists, employees, and customers. Record the answers and their sources. Fix factual gaps first, then improve the narrative. Repeat quarterly. The sophistication lies in consistency, not theater.
What Vectors actually sells
The business model is bespoke professional services. Some clients need a diagnostic, a campaign, a site, or a piece of research. Others retain the agency for continuous monitoring, executive communication, content production, and crisis readiness. Its public materials do not turn this into packaged software pricing. That is sensible: a multilingual watch operation running nights and weekends is not the same product as a board workshop.
Its market position sits between a corporate communications consultancy, a PR and public-affairs shop, a social-listening team, and a digital studio. Bigger networks can offer more geography. Narrower specialists can claim deeper purity in one discipline. Vectors’ argument is that the seams are the risk. If the monitoring team spots a signal that the executive writers cannot interpret, or the data team produces an insight that never reaches the website, the company has purchased expertise without gaining coherence.
There is a pleasant irony here. A firm born when online reputation seemed editable now makes its strongest case by showing how little any single actor controls. Models scrape, rank, compress, and misremember. Communities develop their own language. Employees are broadcasters. Old pages remain alive. Vectors cannot promise command of that system, and it should not. Its more credible offer is attention: enough human, strategic, and technical attention to notice what story is being assembled while there is still time to add better evidence.