Banking
Retail communications on regulated products (MiFID II) are checked before publication, brand voice is monitored across the bank's public channels, and campaign outcome summaries are generated for acquisition campaigns.
See the industry →Polyant is the software for building AI agents inside the marketing team's workflows. Monday editorial brief based on customer conversations from the day before, brand voice covered on every piece of content pre-publication, campaign outcome summary with visible anomalies.
What changes is the starting point of the editorial plan: from the marketing team's assumptions to the real customer conversations of the last 24-72 hours. What changes is the moment of brand voice control: pre-publication, on every piece of content going out, not by sampling once a month. What changes is the timing of the campaign summary: within a few hours of the campaign closing, not weeks later. What changes is how regulated-industry campaigns go out: with mandatory disclosures already integrated, not with the rush to the compliance officer after the creative is done.
These are changes that become visible in the rhythm of the editorial week, not in quarterly reports.
For a large company's marketing team, a squad of agents works outside operating hours too. Each agent does its part and hands the work off to the next, under the same configuration and the same audit registry.
Overnight summary of customer conversations (CRM, ticketing, social) for the week's editorial plan: the brief is ready in the marketing Slack channel before the Monday meeting.
See the agentBrand tone check on every piece of content pre-publication (email, social, brochures), with detection of deviations from the declared register and concrete rewording proposals.
See the agentCross-channel summary of customer feedback (chat, email, reviews, social) aggregated by product area, to bring to the editorial table the questions customers are actually asking.
See the agentPush summary at end of campaign or weekly with anomalies against the baseline: the event stays in the runtime audit registry, readable by the head of marketing with a standard SQL client.
See the agentFor a large company's marketing team, a squad of agents works outside operating hours too and delivers the summary in the morning.
For the head of content marketing at a corporate bank, the Monday meeting starts with the brief already on Slack. Overnight Editorial Brief worked through the night: it read the weekend CRM conversations, the customer service tickets, and the mentions on social channels, and prepared a map of the week's top themes with concrete quotable examples and direct source links.
The marketing team reviews the brief, confirms three priority themes, and adds one of its own. The decision stays with the team. Overnight Editorial Brief collects the confirmation, writes the week's editorial plan into the project management tool, and notifies each owner in their channel of the piece assigned to them.
A copywriter marks a blog article as ready for publication on the CMS. The webhook activates Brand Voice Monitor, which reads the text, identifies two deviations from the declared brand register (use of the superlative "revolutionary", generic call-to-action "learn more"), and proposes concrete alternatives in line with the approved glossary. The copywriter applies both, the article goes to publication.
The four marketing agents interact with different systems — Slack and CRM for the brief, CMS and webhook for brand voice, campaign dashboard and audit registry for the outcome reviewer — but share work state through shared memory and follow the same instance configuration. This is an operating model where specialised agents work together under the same rules and the same audit registry.
In each industry, the editorial side has different constraints and relevant agents. The regulatory detail and the specific cases live on the industry pages.
Retail communications on regulated products (MiFID II) are checked before publication, brand voice is monitored across the bank's public channels, and campaign outcome summaries are generated for acquisition campaigns.
See the industry →Life and health policy campaigns with IDD disclosures integrated from the creative onward, scan of the carrier's editorial content, overnight brief for technical marketing to registered intermediaries.
See the industry →Patient communications with GDPR art. 9 coverage, scan of pharmaceutical editorial content (national pharma code), brand voice of institutional marketing covered pre-publication.
See the industry →Other regulated industries
Utility, manufacturing, public sector, and other regulated contexts have their own editorial specificities: discovery is used to understand where the customer's constraints lie and to build the agent for the industry in which marketing actually operates.
The native channels of the runtime are Telegram, Slack, WhatsApp, and OpenAI-compatible HTTP. Integration with CMS (Adobe Experience Manager, Sitecore, Contentful), email marketing and marketing automation (HubSpot native; Adobe Marketo Engage, Salesforce Marketing Cloud via delivery), social schedulers (Sprinklr, Khoros, Adobe Experience Cloud via delivery) is delivered during the project by the Exelab team.
The typical timeline for a first marketing agent is 4–8 weeks: discovery one week, declarative rule configuration (brand voice guide, industry-regulated claims, campaign baseline) 2–3 weeks, integration with CMS and marketing automation 1–2 weeks, handoff to the marketing team one week. The duration depends on the number of integrated systems.
Operational management of the agents — including brand voice rule updates, editorial taxonomy changes, and campaign baseline adjustments — stays with the customer's team via the graphical admin panel or directly in TypeScript. For those who want to delegate operational management as well, Exelab provides managed services in three profiles: Core, Advanced, and Ultra. Details are available on /managed.
The value of marketing agents is measured across three types of outcome. Time reclaimed by the marketing team compared to the pre-agent baseline (editorial brief preparation, brand voice review, campaign outcome summary). Reduction of editorial compliance errors (non-compliant claims published, missing disclosures). Speed of response to campaign anomalies (days ahead of the manual pattern). The baseline is established during discovery and measured in the first weeks of operation.
A 30–45 minute conversation to understand how Polyant for Marketing applies to the customer's situation and how long it takes to get the first agent into production.