Programmatic Media Explained

causaLens x Syneos Health: Programmatic Media Explained. Speakers: Oliver Matthews (causaLens), Vikeesh Velumalan, Director of AI Strategy and Innovation (Syneos Health).

Oliver Matthews: My name is Oli Matthews. I'm part of the go-to-market team at causaLens, and we're working very closely with Syneos Health to help them become AI native across the full commercial pharma workflow. Today I'm joined by Vikeesh from Syneos, who I work with very closely. Vikeesh, it would be great if you could give a quick intro.

Vikeesh Velumalan: Thanks, Oli, and thanks everyone for joining. I'm Vikeesh Velumalan, Director of AI Strategy and Innovation at Syneos Health. I sit on the media side of our business, and my role is to work out where we can integrate AI meaningfully into our media offering and services. I identify and implement opportunities to improve how we plan, buy and optimise media, as well as media data integration for our pharmaceutical clients. Previously I was a Director of Media Solutions and one of the leads responsible for building our programmatic practice at Syneos Health.

Oliver Matthews: Thanks, Vikeesh, and thanks for joining us today. causaLens and Syneos have been collaborating over the last couple of years, implementing Digital Workers and agentic AI solutions across the full span of commercial pharma operations. Over the next 30 minutes, we want to bring to life a very real example of how Syneos Health has adopted AI within programmatic media: how Syneos decided programmatic media was a good area to rethink with AI, what the investment means for the pharmaceutical clients it serves, and how AI is changing the way the team operates. To get started, Vikeesh, could you tell us a bit about the programmatic media team at Syneos Health, what you do day to day, and what the operations look like?

Vikeesh Velumalan: Sure. We run an in-house media practice out of Evolvix, the media wing of Syneos Health Communications, and we have a dedicated and seasoned trading desk. Day to day, the programmatic team covers the full life cycle of a programmatic campaign: media strategy, planning, setup, QA, activation, optimisation and reporting.

At Syneos we're very data-driven, so more often than not our process starts with curating the right audience. In pharma, that's generally the healthcare professional, where we're building and refining physician targeting down to the individual provider level. We also look at reaching patient and consumer audiences effectively, with HIPAA compliance fully in mind. Once we've determined the audience, we look at which channels and partners are best suited. Then comes campaign build and trafficking, and creative quality assurance, and that completes the setup process.

The part that never really stops is monitoring in-flight campaigns, observing pacing and delivery, and making optimisation decisions. Given the number of activations and KPIs, this can get very extensive. The two things that make programmatic media in pharma distinctive are scale and regulation. We're running a lot of campaigns concurrently across multiple demand-side platforms in a very fragmented healthcare publisher landscape, and every one of those decisions sits inside a regulated environment where data privacy and compliance dictate the operation from the very start. Thankfully, we have a great team of specialists doing genuinely expert work.

Oliver Matthews: Very cool. So it's quite an involved set of activities, from campaign setup through to in-flight monitoring and optimisation, with all the compliance and privacy challenges that come alongside. What did this workflow look like before you started bringing AI into it, and where was a lot of the manual and repetitive effort going?

Vikeesh Velumalan: Honestly, a lot of it was fairly labour-intensive given our soup-to-nuts approach. We had a refined workflow in place thanks to our team's experience, but that doesn't change the time and effort some of these tasks require. Building out campaigns line by line, then doing it again for the next market, the next audience, the next flight. Trafficking, tagging, checking creative, and making sure we meet high standards of accuracy and QA is a significant amount of manual effort.

There was also a lot of effort spent gathering data: logging into multiple platforms, exporting, stitching it together, and getting it into a format someone could look at and form a view. Monitoring was similar. It's not set and forget. It requires daily review to make sure everything is going to plan, and troubleshooting when it isn't, which can take quite a bit of time, especially for a novel issue. When I evaluated where AI could drive the most impact in our workflow, it was very evident: the mundane, repetitive tasks. We're comfortable and confident those should be taken off our experts' plates so they can focus on their craft.

Oliver Matthews: Amazing. It's exciting for the team to get back to the more interesting parts of the workflow, the strategic thinking about how to optimise campaigns, instead of spending a lot of time reviewing creatives and setting up campaigns, work that AI is a very good fit for. What made you decide to take this AI-driven approach? Was there a particular trigger, or was it the weight of manual processes building up within the team?

Vikeesh Velumalan: There wasn't one dramatic moment. What's great about Syneos is that we're constantly looking forward and believe in being early adopters of best practice. When Adam Cahill, who leads Kinetic at Syneos and was my lead at the time, was looking for opportunities to bring the benefits of AI into our business and operations, I raised my hand and explained why programmatic, given how it's structured and operates, should be first on the list.

The team is small and nimble, and volume was growing. There was more demand for our services, so the business case started to write itself. Client and internal stakeholder expectations have grown, and we have an opportunity to meet them in a far better way than before. As an organisation, we can see that the AI age applies to everyone, and personally I think not finding ways to transform would be a disservice. If we can adopt AI in a way that empowers our team to do meaningful and satisfying work, reduces friction, and improves our service levels and general capabilities, it's a no-brainer. Give the team their time back and give them a better basis for decision-making. That's why a partnership like the one with causaLens makes sense.

Oliver Matthews: It's great that we took a business case-driven, outcome-driven approach, and that scaling existing activities to more client work offered significant value to Syneos. But beyond scale, there's also the opportunity to improve the decisions themselves. So there's a scale benefit and a decision-making benefit in this workflow. Where has AI created the most value across both?

Vikeesh Velumalan: I'd say three areas: setup, monitoring, and data synthesis. The first is the least fun and probably the most valuable: quality assurance, the repetitive build work, the checking, the reconciliation. It's time-intensive and it's where errors can creep in. Taking that off people's plates is an immediate and visible win, and it's the kind of thing that gets the team on your side as you implement this.

The second is monitoring. We're a platform-agnostic shop, monitoring every campaign continuously across different platforms. Solving that with automation and AI has improved our workflow and flags both positive and negative patterns early. That shortens the gap between something going off track and someone knowing about it, which helps our performance.

The third is synthesis and explanation: bringing signals together across platforms and channels and saying, "Here's what appears to be driving this." Our team can often make the connection anecdotally, but having data behind it gives more confidence. That drives better outcomes, and clients see the value of the programme beyond media ROI. And it compounds. The time saved on the first two creates the capacity to act on the third. If you only automate and don't improve decision-making, you've just made the same decision, faster.

Oliver Matthews: It's been great to go on this journey with you and the team and realise the value AI can bring. But we've heard a lot of concerns in the media recently about AI, particularly at the frontier, and its potential downsides. How are you thinking about mitigating those risks at Syneos?

Vikeesh Velumalan: Most of those concerns are reasonable and worthy of healthy debate among experts and leaders in the field. We have experts in AI governance at Syneos, and I'd defer to them on AI regulation specifically. My view is that, at a minimum, there needs to be an appropriate framework that implements human-in-the-loop. I want to be precise about what that means, because it's become a phrase people say to sound responsible. For me, it means a person approves anything that moves money, changes targeting, or reaches a potential patient or physician. The system does the analysis and makes the case, but a human gets the final word.

In the projects I've worked on with you, Oli, we've been very intentional about this: tightly scoped use cases rather than a general-purpose approach, clear rules about what data goes near a model and what absolutely does not, a requirement that any recommendation shows its reasoning and the data behind it, and an audit trail so that months later we can reconstruct why a decision was made and who made it. That's imperative.

Oliver Matthews: That makes a lot of sense, and it's something the causaLens platform does very well, capturing the audit trail behind our Digital Workers. We can easily understand the work they've delivered and use it to understand past decisions, and the human oversight is captured in those audit logs too. Because you work in such a highly regulated industry, are there extra considerations you need to think about?

Vikeesh Velumalan: Definitely, simply because the cost of being wrong is asymmetric. In most categories, a bad optimisation might waste some money. In ours, it can create a compliance issue or reach someone in a way that isn't appropriate. For as long as I've been on this team, we've been very strong and intentional about mitigating these risks. That takes sound judgement and context, which AI doesn't always have, at least not yet. Someone could make an announcement tomorrow that makes this moot, but even then, we would still validate an AI's judgement here. It takes a human years of training to be considered qualified and trusted with a complex or risky task. We should apply similar standards of vetting and validation based on the stakes.

Healthcare media has constraints a model has no natural awareness of. Approved messaging that has been through medical, legal and regulatory review can't be altered on the fly, and AI won't necessarily know that. Privacy rules govern what you can infer about a person, not just what you can store. And with health conditions, efficient targeting and appropriate targeting are often not the same thing. An AI engine may run straight past that, because it isn't the objective of its function, until someone tells it not to.

The other piece is accountability. If a client or regulator asks why a decision was made, nobody will accept "the model decided" as the answer. Somebody has to be able to stand behind it. So a qualified, context-rich human has to be part of decision-making.

Oliver Matthews: That accountability piece is key, and very relevant in healthcare and pharma. It's been great working with you as we've built those checks and balances into the Digital Worker workflows we've defined. Human oversight is super important, but we're moving to a point where we've implemented AI-generated recommendations and may move on to AI-generated actions within the programmatic workflow. What have you learned about building trust in AI recommendations, and ultimately trusting AI to take actions?

Vikeesh Velumalan: I don't know about you, but I don't always accept Google Maps rerouting me. Often it's because I've driven in that neighbourhood before, so I have context and can be critical of the recommendation. It's generally accurate, but I know it isn't factoring in everything, like how dangerous a left turn is on the new route, or that school's back in session and kids will be congregating around an intersection.

The same applies here. Not every recommendation, even one grounded in truth, needs to be acted on. It still takes someone with tacit knowledge to understand the impact. In media, a tactic in a plan may have a lower conversion rate, but its objective in the programme is to expand reach. Both are KPIs, but only one is really qualifying that tactic.

Accuracy on its own doesn't earn trust. The system has to show its working so a subject matter expert can follow the logic and interrogate it. It has to be grounded in data the team recognises, and it has to arrive where they already work, not in another tab someone has to remember to open. Reducing friction and making it easy to use is a big factor in implementing new workflows. You also have to make it easy to disagree, and capture it when people do. A system that can be overruled, and learns from being overruled, gets trusted faster than one that presents itself as always right. That feedback loop has been instrumental in refining the quality of our output.

Oliver Matthews: The Google Maps example brings it to life nicely: a highly useful system people rely on every day, but not one you blindly trust to act on your behalf. It's been great to see the progress we've made together designing these workflows and bringing AI into the programmatic media process. If another organisation is thinking about starting this, or has attempted it, what would you say they need in place before getting started?

Vikeesh Velumalan: The best place to start is being honest that it's not just a technology project. A lot of organisations are excited about AI, rightly so, but the real challenge is turning that interest into something practical, compliant and useful in the day-to-day reality of healthcare media. That's usually where we can help at Syneos. We understand the ambition, and we also understand the constraints: data foundations, workflow, governance, privacy, MLR, media performance, and keeping people meaningfully in the loop. Those pieces have to come together for AI to move beyond a pilot or proof of concept.

If a client or partner is thinking about where to begin, I'd say reach out before the use case is fully defined. That's often the most valuable time to have the conversation. We can help pressure test where AI makes sense and where it may not, what needs to be in place first, and how to design something measurable, responsible and grounded in how healthcare media really operates. The goal isn't to sell AI for the sake of AI. It's to help teams find practical opportunities where the technology reduces friction, improves decision-making and creates value without compromising the standards this category requires.

Oliver Matthews: It's been great to help the Syneos team identify those opportunities within programmatic media and rethink the way the work gets done. We're not finished yet, and we're looking forward to continuing to work together. Vikeesh, where do you see this taking us next, whether that's greater automation, better decisioning, or other opportunities in programmatic media?

Vikeesh Velumalan: It's difficult to predict the exact timeline in a space evolving this quickly. The technology is moving fast, the platforms are changing, and comfort levels differ across organisations. So rather than a big leap into automation, I think the right approach is a crawl, walk, run progression.

The crawl stage is about assistance: taking manual work off the team, helping them see patterns faster, and giving them better inputs for the decisions they already make. The walk stage is where the system starts making recommendations with clear reasoning and evidence behind them, so the team can build confidence. The run stage is where you may introduce more supervised or bounded automation, but only where the process, governance and measurement are mature enough to support it.

That's the future state we're building towards at Syneos. Not automation for its own sake, but smarter campaign operations where repetitive and time-sensitive work is handled more efficiently, and our people can focus on strategy, judgement and client outcomes. In healthcare media especially, that balance matters. You want to move quickly, but also very responsibly.

For clients and partners, the opportunity isn't to have a perfect AI roadmap figured out. It's to have the right conversations early. Which parts of the workflow are creating friction? Where would better decision support help? What can be safely automated, and what should remain firmly human-led? So my open invitation is: if you're trying to understand where AI can create practical value in your media operations, reach out to me and the team at Syneos. We're already working through these questions in healthcare media environments and can help you think through where to start.

Oliver Matthews: Thanks, Vikeesh, and thanks everyone for joining today's discussion on how Syneos Health has rethought its use of AI within programmatic media. It's been great working with Syneos on this project and others, and we're excited to keep collaborating across the full span of commercial pharma workflows. As Vikeesh mentioned, please reach out to either of us to learn more about how we're applying AI across commercial pharma. Vikeesh, thank you very much for your time.

Vikeesh Velumalan: Thanks, Oli. This has been great, and it's always fun working with you.

Oliver Matthews: Thanks everyone. We'll be continuing our webinar series on our work with Syneos shortly, so stay tuned for the next one.