Over two days this October, Brightspot leaders, customers and partners worked through one question. If AI makes content cheap to create, what does it take to run it at machine speed without giving up judgment or trust?
Brightspot CEO Raleigh McClayton opened Illuminate 2026 with a question he said he can’t stop thinking about: “If creating content has never been easier, why does running your content feel harder than ever?”
Then he put numbers on it. By 2030, 90% of net new content on the internet will be synthetically generated or AI-augmented. A team producing 13 pieces of content a week today could be producing about 1,000 a day.
For 30 years, every content system was built on the assumption that content is expensive and slow to make. “Including ours,” Raleigh said. AI flipped that. His read: creation is solved, and the next decade belongs to teams that get good at orchestration, curation and trust.
Pretty much every session that followed came back to one of those, whether the speaker was reviewing a pull request or explaining why a photo of a sea turtle needs alt text. Here’s how they connected, plus 10 takeaways to bring back to your team.
When everyone can produce anything, abundance becomes worthless and judgment becomes priceless. The next competitive skill is knowing what not to publish.
Orchestration: From “can we make it?” to “can we move it?”
Content used to travel down a line of people. Now, Raleigh said, it moves through a spiderweb of people and agents, across more channels than any team can manage by hand. The job is getting the right asset, in the right form, to the right place, with the right approvals, every time.
“That is an air traffic control problem at a busy airport,” was the analogy used by Raleigh in his opening keynote.
He calls this the shift from content management to content operations, where the question for the next decade is how your content operates. (Raleigh’s name for his own industry: “the category formerly known as CMS.”)
Brightspot’s answer to the air traffic problem got the most airtime: Esca Automations, a standalone drag-and-drop automation product that connects to every Brightspot environment and the rest of your stack, including Slack, Teams, GitHub, Jira and any system with an OpenAPI spec or MCP server. It’s available now.
Later, in their sessions covering practical uses of agentic AI, Lucy Collins (Director, Customer Engagement) and Will Chu (Senior Vice President, Customer Engagement & Product Delivery) walked through three automations built from real customer conversations:
- An AI quality check that catches typos, broken links, missing tags and leftover placeholder notes (including inside shared snippets) before an article reaches a human editor.
- A marketing brief that becomes two draft landing pages, built only from the theme’s approved modules, so editors never start from a blank canvas.
- A daily 8 a.m. check of an EPA RSS feed that found seven new entries, created a pitch with seven linked assignments in Assignment Desk and posted a summary to Slack.
When they polled the audience on which stage of the content lifecycle eats the most time, reviewing and approving came out on top. Will pointed at the handoffs: the Slack message, the copy-paste into a project tool, the reassignment, the follow-up two days later. Automation should “stop requiring people to be the glue,” he said.
Tom Neff, director of product on Brightspot’s platform team, gave the clearest example in another session showcasing the new and different applications of agentic AI. A company rebrands, or a fact repeated across hundreds of articles turns out to be wrong. That used to mean an editor opening articles one at a time, or a ticket sitting in a developer’s queue. With Esca, the ticket itself starts the workflow. An agent finds and updates every affected item, an editor approves the changes in Slack, and the results post back to the ticket.
“Work that would have consumed an entire afternoon now only takes a few minutes,” Tom said.
Curation: Judgment becomes the scarce resource
When everyone can produce anything, CEO Raleigh McClayton said, “abundance becomes worthless and judgment becomes priceless.” The sharpest line of his keynote was about restraint: “The next competitive skill is knowing what not to publish.”
For the editors listening, he said this is the decade their judgment “finally gets valued properly and recognized, applied at scale instead of spent on busy work.”
Our commitment to you is that we will build tools that ensure your judgment is never secondary.
The principal of careful, human-in-the-loop review and oversight shone clearly through Will and Lucy’s deep dive on agentic AI in action.
Take the example of an AI automation to ensure brand styles are being met. The agent edits with track changes on and adds inline comments when it’s unsure, so an editor accepts or rejects each suggestion. Playing a customer during one of their role-play discovery calls, Will summed up where a lot of teams sit with AI right now: “I don’t know if we wanted to do the full mile, right? Maybe we’ll do a couple of kilometers.”
Brightspot CTO Lee Teague applied the same thinking to engineering. Most of the code he writes now comes from an agent. He walked through a real feature (an agenda module for this very event) going from requirements to production with Claude Code, automated code checks, security scans, peer review and an AI review step in GitHub Actions.
His instruction to developers is to “take the same time to deliver a higher quality product.” That means more unit tests and deeper reviews. Writing code is maybe 20% of a developer’s week, he pointed out. Writing it twice as fast doesn’t turn a 40-hour week into an 80-hour one.
Judgment matters in search, too. In the AEO session, Miles De Feyter from Brightspot Creative Services was asked for the one must-do for AI visibility. “You need to have a unique viewpoint,” he said. “You need to be talking about and presenting something that is worth citing.” A generic article with no stance or data behind it, he added, “is empty and it’s useless.”
Choosing a CMS for editorial workflows? Ask who owns the review
Trust: Governance is the moat
Raleigh’s third challenge was trust. Within a few years, most of what’s published online will have been made at least partly by a machine. In that world, “integrity becomes brand equity,” he said.
Governance, verification and a single source of truth used to sit in the compliance column. “That is the moat,” Raleigh said. He expects an integrity premium to show up in search traffic, AI answers and customers’ willingness to act on what you publish.
Agents in Brightspot never get more authority than the user behind them. “If anything, they have less,” Ravi said during the leadership ‘Ask Me Anything’ session. MCP access runs through OAuth on the user’s own permissions, and reading content is a separate permission from changing or publishing it.
Esca logs every run, down to what changed and how many tokens the AI used, and those logs can feed an agentc Security Information and Event Management (SIEM) platform. The right to build an automation is also separate from the right to switch it on. “We imagine that the person who builds an automation won’t necessarily be the same person who enables it,” Tom said.
During his guest-speaker session covering “The liquid content era”, the International News Media Association’s (INMA) Robert Whitehead, digital platform initiative lead at INMA, took trust somewhere less comfortable. Trust in news media has fallen to 37% in many markets, and 2026 is the first year more people get their news from platforms and creators than from news organizations.
Telling people to trust you because your brand is 200 years old, or because the awards are on the shelf, hasn’t moved the needle. “Trust is not a strategic moat for news media companies,” he said.
Robert’s prescription is trust hygiene: showing how the work gets done, who reported a story and why, with some of the rawness creators bring. “Sometimes raw is better than highly refined.” Rigor still pays, he added: AI companies need fact-checked content, and they’re starting to pay for it.
Illuminate’s own Ask Me Anything with Brightspot’s leadership team ended on the question underneath all of this. An attendee asked about a CEO who reportedly dropped a $600,000-a-year Salesforce contract after vibe coding a CRM in two months.
CTO Lee Teague’s answer was buyer beware, since you’ll be supporting it yourself: “Your vendor is you.” Ravi Singh went further.
“An Excel spreadsheet isn’t a CRM. You can’t run your business on an Excel spreadsheet,” he cautioned.
People come to Brightspot for the trusted content infrastructure you don’t want to rebuild, Ravi said: identity, permissions, content provenance, audit trails, versioning, workflow reliability, accessibility and security at enterprise scale.
10 questions before AI enters your governance framework
10 takeaways from Illuminate 2026
1. Start with the work, then decide where AI fits
Every automation Lucy and Will showed began with a business problem: inconsistent review, the blank page, scattered monitoring. Their tip for finding candidates: listen for the tasks your teams already complain about.
2. Don’t fire the AI bazooka
Dan Slaughter, Brightspot’s chief architect, borrowed a colleague’s term for automations that hand every step to a model. Use AI for the steps that need it and keep the rest deterministic. It’s easier to audit, and you stop burning tokens on work a script can do.
3. Requirements are the new bottleneck
When Lee polled the audience on their biggest development bottleneck, the top answer was gathering and defining clear requirements. Brightspot Atlas, an index of all Brightspot code with an MCP server, helps agents follow platform best practices. It’s internal today, with customer availability targeted for late Q1 2027.
4. Configure first, integrate second, write custom code last
Ravi’s order of operations for extending Brightspot. Lee added a related nudge: stay on the upgrade program, because if your version lags, so do your AI capabilities.
5. Treat AI crawler access as a business decision
Cloudflare now splits AI bot permissions 3 ways: search, agent and training. But Googlebot handles both search and training, Miles noted, so a blanket “block AI” setting can quietly cost you search visibility.
6. Spread your citations across platforms
On August 8, ChatGPT changed how often its fan-out queries searched specific sites. 6 days later, Reddit’s share of ChatGPT citations had dropped 86%. Most of what Reddit lost went to official brand sites and documentation. Build citation breadth across your site, docs and trusted publications.
7. Measure two or three things that drive decisions
Miles’ advice for a sea of AEO tools, from share of model to Bing’s free AI performance report. Brightspot tracks about 20 topics and 250 prompts in Profound. Since May, its visibility rose from 13% to 21% and brightspot.com climbed from the 7th to the 2nd most cited domain in its category. (The category leader still shows up in 69% of answers, so there’s work left.) Changes now register in about a week.
8. Most accessibility fixes are one missing field
Basima Zafar, a senior software engineer on the Brightspot Toolkit, watched her mother, who uses the JAWS screen reader, hear “link, link, link” on pages where nothing had a name. “Nearly every one of these problems is missing an attribute somebody didn’t fill in,” she said. As Lee noted in the AMA, well-formed, accessible HTML is also what AI agents parse best.
9. Fix problems where you spot them
The Toolkit browser extension began as one customer’s homemade “find this page in Brightspot” button. About 500 editors now use it to catch issues like long title tags and missing alt text on the live page, then jump straight to the item in the CMS. On Sid Jennings’ demo page, only 1 of 12 images had alt text. AI-generated alt text is coming soon.
10. Long-form video is the most flexible source format
Robert Whitehead’s case for “liquid content”: agentic AI can reshape one story by format, length and tone for each audience. Long-form video converts most easily because it comes with a script, images and audio.
