Seven lessons from onboarding Agent Platform in a regulated business

Optimizely’s 2026 European user conference Opticon showcased the vendor’s vision for how Agent Platform can support digital and marketing teams to adopt AI in a meaningful way. As part of the event, UNRVLD presented lessons from onboarding Agent Platform in a regulated business alongside one of our global financial services clients.

100 days into adopting the AI orchestration platform across the client’s marketing and digital functions, it was the perfect time for us to provide an honest appraisal of what they had achieved so far, what had worked well and what we would change next time.

During the session, our Head of Strategic Advisory Services Dom Graveson advised on realistic goals for the first 100 days of adopting the platform. We outline these below, alongside seven lessons to set you up for success while planning your Agent Platform onboarding programme.

Realistic aims for your first 100 days adopting Agent Platform

There are five dimensions to consider when planning to adopt any AI orchestration platform, and it is helpful to have a tangible goal for each that you can strive for within 100 days. These goals support each other and will leave you in a position to move forwards with your deployment and a plan for what should come next.

People – Users engaged, trained and actively experimenting

Organisation – Ownership, accountability and decision-making clearly established

Processes – Workflows tested and measurable improvements demonstrated

Information – Data sources, permissions and gaps fully understood

Technology – Experiments validated, limitations understood and next steps identified

Our client has been able to work towards these goals in just a few months, despite the additional constraints of working in a regulated industry, with the seven lessons we quickly identified during the onboarding process.

Starting small is better than not starting

Adopting a now, next and later framework immediately brings some strategic order to the adoption of an AI orchestration platform, without needing to fully plan your roadmap for the next year or more.

Start small by building some proof of concepts to test and learn from before you commit to more complex workflows that require access to data and systems outside of your team’s control. This allows you to understand what you do and don’t need in order to keep up momentum in your adoption plan, what should come next and what should be pushed further out on your roadmap.

Solve simple, specific pain points to test and learn

The ideal subjects for proof of concepts should be processes that are completely in your control where the output can be easily assessed for quality and consistency. Ideally it should be a regularly repeated process that is identified as a pain point within the team because it is time-consuming but necessary.

Such pain points are the easiest to optimise the output of and learn from. Your team will also be motivated to support the automation of the process.

Solve for team-wide issues, alongside personal productivity

We have worked with our financial services client to build and experiment with several proof of concept multi-agent flows, to automate business processes related to content production and updates for their website. Both the initial POC subjects had been identified as pain points by the team.

These proof of concepts help to demonstrate the scale of efficiencies that can be unlocked across the team through automation of repeatable time-consuming tasks, as well as agent engineering, data access and governance that is required to deliver a fully-functioning solution.

Connect the systems you control at first to prevent blockers

For marketing and digital teams, initially connecting systems that you own and control prevents blockers from corporate IT teams and company-wide policies from impacting your immediate progress.

Connecting your CMS, website analytics etc will limit the agents you can build, but allows you to build confidence and trust in the system through POCs before permissions are sought for connections to other core business systems.

Allow yourself time to play to understand what’s possible

One of our client’s team members joined our June 2026 AI Workshop as a starting point in their adoption journey with the goal of building an agent in a morning. The idea she conceived during that exploration session was expanded out to become one of the first proof of concepts we supported.

“Playing” with the platform will help your team to brainstorm what might be most beneficial, and it is important to keep on playing as AI platforms and LLMs continue to evolve at pace.

Build capability and confidence early

Consider your people within your adoption plan. They will find it much easier to conceive ideas, test them and learn if they have developed confidence and capability.

Optimizely offers a five-day live training programme called Markademy to offer immersion and quickly build capability and confidence.

AI is a journey, not a destination

Developments that have occurred in the past year alone demonstrate how quickly the possibilities of AI are changing, and how relatively early we are in the adoption cycle. Much like the continued evolution of digital products and experiences, there will always be a next stage in the development of AI.

We don’t recommend that our clients relaunch a website and then don’t touch it again for two years, and believe the same is true of AI.

When we talk about supporting clients to harness the Intelligent Future, we are describing a new era, not a single moment in time. For clients who have invested in Optimizely’s digital and marketing technology, adopting its Agent Platform is one way to begin laying foundations on which to build in that new era of intelligent experiences.

Get in touch if you would like to chat to one of our experts about what the first 100 days of adopting AI within your digital or marketing function could look like.

And if you already leverage Optimizely’s technology, you are welcome to sign up to our next half-day AI Workshop on November 18th, where you’ll be able to build an AI agent alongside a group of your peers supported by the UNRVLD team.