The market is changing fast. New models ship every few days, and each one is measurably smarter than the last. At Airtable, we are continuing to build what's next and our strategy is grounded where it always has been: in what our customers are already building.
As I talk to companies attempting to put agents into real operational work, they keep circling back to the same question: if the models are smart enough, why are the outputs still lacking?
A recent VentureBeat survey found that 57% of enterprises have traced a wrong agent answer back to missing or inconsistent business context. In our experience building agents at Airtable, we've found something similar: the intelligence is there, but the outputs fail when the agent doesn't have the right context.
Most agents build their own context graph by reading whatever's available: emails, docs, Slack, meeting notes. The approach is fast, and it feels thorough. But it's assembled from people on your team describing the business a dozen different ways, none of it reconciled, prone to error and inconsistency.
Some companies figured this out early and took a different approach. Our most successful customers don't allow agents to guess at the shape of their business. Instead, they put agents to work in a structured, relational system where humans already operate — tracking roadmaps, managing campaigns, running feedback pipelines, and coordinating releases.
The impact is more than accuracy, it's immediacy. When an agent mines Salesforce or Jira, for example, it doesn't file the result away for someone to search for later. By the time a product manager or campaign lead sits down in the morning, the information is current and the workflow's already moving — their job is to make the work better, not go find it first.
With this in mind, here’s how teams across several industries, including media, healthcare, publishing, and retail, are putting agents to work, and what it's teaching us about what they need to succeed.
A major global music company
This company is one of the world's largest music companies, operating labels across every genre and geography. Each label historically ran its own roster, release calendar, and way of working. The team consolidated all of that data into a single foundation in Airtable — then put agents to work.
How humans and agents work together: The moment a new artist contract is uploaded, agents parse it and extract the terms that carry operational weight: royalty rates, creative-support clauses, licensing approvals, name-and-likeness terms. The agents are chained, so one's output feeds directly into the next. Terms then route automatically to the team responsible for taking action — royalty rates to finance, licensing approvals to sync, video commitments to marketing.
The result: The team is saving hundreds of thousands of dollars annually from consolidating separate software licenses. What had been dozens of disconnected tools across 59 labels is now one unified system that extends into 70+ locally-tailored apps, serving every team from artist relations to licensing. The whole system was deployed in roughly ten months.
"Before we could build the AI tools, we had to fix our foundation. We needed a baseline of truth that everyone could go to and trust."
A healthtech company advancing cancer care
This company uses real-world evidence to advance care and research across hundreds of cancer centers nationwide. The company runs dozens of product teams and more than fifty roadmaps. Before agents could be useful, product operations had to connect portfolio management, customer feedback, competitive intelligence, and market insights in one structured system in Airtable.
How humans and agents work together: Every piece of incoming feedback gets read against the same fixed set of questions: what's the core problem? Is there a feature request? Has this come up before in the last 90 days? The agent’s answers write back as structured fields on the same record a product manager already has open.
A separate agent connects the company's clinical data to its own sales records, matching thousands of trials and drugs with the biopharma companies developing cancer treatments. When that connection lives in the data instead of in emails and decks, building an account plan for a potential customer takes half an hour instead of half a day.
The result: Feedback processing that used to take four people four weeks now takes one person one week. A data-connection project originally scoped at a year of engineering work shipped in two weeks.
"By the time a product manager opens a request, the homework is already done."
One of the world's largest news publishers
This publisher produces 5,000+ stories a week across newsrooms on multiple continents. When the editorial team needed to bring structure to how stories were planned and tracked, they mandated that every story starts in Airtable.
How humans and agents work together: Every published story flows through a single automation. An agent reads the headline and body copy, checks facts, grammar, and house style, and writes the assessment back to the record. Clean stories get a tick; problem stories get a targeted note. Findings are immediately distributed to the assigning editor, reporter, and news desk among hundreds of Slack channels globally — all within minutes of publishing.
The result: 5,000 stories a week now flow through the system, checked by AI, for more than 1,000 journalists across newsrooms on three continents. The rollout went from a single go-live to every global office in under four months.
"The importance of having the human in the loop here means it gives confidence to journalists that these are the things they should be checking."
A high-growth home design brand
This brand makes customizable lighting and hardware. It's eight consecutive quarters into double-digit growth, and the entire copy team is two people. When the team moved its product-launch process into Airtable, they taught agents the brand voice well enough to trust them with the first draft.
How humans and agents work together: Six agents generate product copy, including names, descriptions, selling points, and specs. Each pulls from curated language libraries and brand-voice examples. Evaluation agents score outputs against guidelines, and a separate agent screens for regulatory and standards risks. Approved copy flows into an HTML-ready field, and every draft gets a human edit before it reaches the live page.
The result: More than 600 products launched in a single season, with a copy team of two. A pilot launch shipped with 100% AI-drafted, human-edited product-page copy.
"We're not just making AI slop here. We need our agents to operate at a higher level. We want them to feel polished, inviting, and to carry design authority."
Every company above started the same way: they built the system before they deployed the agent. The real divide opening up right now is not who's using AI, but whose agents are working from a real picture of the business instead of a guess.
Airtable crossed 1M standalone product sign-ups last quarter. Most of those teams were already thinking in structured data, automations, and workflows before agents entered the picture. That turned out to be their advantage.
