How Nomination runs millions of AI executions a year, from the edge to the core business
France's B2B data reference, 450,000 decision-makers and 100,000 companies tracked. Rather than hiring an in-house AI team, Nomination chose a hybrid model: its data experts stay on verification, Tandem operates the AI pipelines and agents on n8n on a continuous retainer, at the scale of millions of executions.
Industrialising at the scale of millions, without diluting human quality
Nomination built its reputation on the human verification of every contact, its historical differentiator against automatically scraped databases. But the inbound volume, nomination newsletters, executive moves, company updates, grows faster than the natural capacity of a manual team. The question wasn't to replace the human, but to industrialise everything that could be, upstream, at the scale of several million executions a year.
Decision-makers to keep up to date
An inbound flow (nominations, executive moves, company updates) growing faster than a manual team's capacity.
Of items to process a year
A volume impossible to absorb by hand without industrialising everything that can be, upstream of verification.
Costs to control at scale
At massive volume, every cent per execution counts. The challenge isn't running the AI, but running it at a controlled cost.
Nomination, ParisNomination
Nomination is France's reference player in B2B data. The platform continuously tracks 450,000 decision-makers and 100,000 companies, with human verification of every contact, a strong differentiator against automatically scraped databases.
The data feeds both large-account prospecting tools and the "Appointments" sections of major French media. Maintaining quality at scale is the central challenge of the business.
Start at the edge, industrialise, then take on the core business
Rather than hiring an in-house data engineering team, Nomination runs a hybrid model: its data experts stay on verification and quality, where they create value, and Tandem operates the AI pipelines and agents continuously on retainer. The bet: start on edge processes to prove robustness, then move progressively toward the core business.
Start with edge processes
First end-to-end pipelines on non-critical tasks: database enrichment, news research, identification and matching of nominations. Robustness is proven in production before touching the core business.
Industrialise at scale on n8n
Scaling up to several million executions a year. Systematised error handling, continuous monitoring, pacing of massive batches to hold the load without breaking production.
Control LLM costs
Switching to deterministic building blocks for all simple cases, LLM calls reserved for genuinely ambiguous ones, model choice per use case. Result: an AI bill that stays controlled despite the volume.
Move up to the core business, refocus humans
Progressive extension to core-business processes with specialised AI agents. Internal data teams are refocused on verifying the most important flows, where humans add the most value.
An AI capacity industrialised at scale, with no internal hire
Several million executions a year operated on n8n in continuous production, a series of specialised AI pipelines and agents covering edge to core-business processes, errors kept under control and LLM costs contained despite the volume. Nomination's data experts stay focused on verification, where they create value.
« Our value is data quality and human verification. AI helps us handle the volume without compromising on that standard. »
Want us to map your requests?
Free audit of your processes, we identify high-ROI use cases together, and quantify the gains before deploying anything.