“The true power of AI lies not in replacing humans, but in working alongside us to achieve what neither can do alone”
– AI pioneer Sebastian Thrun.
Whether you love it or hate it, there’s no denying that Artificial Intelligence is here to stay, and evolving at pace. With every passing minute the applications grow wider, the debates grow stronger, and the stakes get higher. The pressure to get involved or get left behind can be overwhelming.
Onclusive has been integrating AI into platforms and workflows for some time now, helping improve efficiency and enhance output, but we’re far from the only ones. More and more businesses are turning to AI to replace existing processes and save costs. In the UK, a Trades Union Congress (TUC) study published in August 2025 found that 51% of the UK public are concerned about the impact on their jobs.
Nowhere is this more prevalent than in the realm of data collection and analysis, where AI has already proved it can do incredible things, which got our Planning team thinking. Justifying their title, the team took upon themselves to get ahead of the million dollar question that was inevitably coming down the tracks:
“Couldn’t I just use AI to do the job of Onclusive Planner?”
Challenge accepted: Player 2 has entered the game
To make it clear, our intention was never to catch AI out, or indeed prove our own worth, but to formulate a genuine response to the question; understand where AI’s strengths may lie in this field; and explore ways Onclusive might be able to use the technology to enhance our own research.
Adopting the roles of both journalist and comms professional, our UK Planner team gave three of the leading public AI engines, Gemini, ChatGPT, and Co-Pilot, the exact same prompts to see how they would fare against our 37-strong Forward Planning service research team. Known for its more human approach, Claude was also included in the research initially, but dropped fairly swiftly for reasons that will become clear later.
The below represents the kind of challenges presented to said AI engines:
- What are the big news events happening next week?
- When would be the best time to put out the attached press release?
- Please create a calendar of noteworthy events for my chosen sector for 2026.
- Which companies in this sector are announcing their results next month?
- And perhaps most challenging:
The attached is a spreadsheet of all the events on Onclusive Planner for today.
Please create an equivalent for tomorrow.
The results were revealing to say the least, but to put them in context:
At the time the bulk of the research was done, Onclusive Planner carried 3,784 events for the coming week (an average of 540 per day), of which 215 were categorized as a Top Story and 100% had been verified by our research team. Listed events included monthly UK inflation figures; a Cobra meeting on Iran; the BAFTA TV nominations; court hearings for both Nicolás Maduro and four 9/11 suspects; the annual World Meteorological Organization State of the Global Climate report; the installation of a new Archbishop of Canterbury; and a speech by Wes Streeting on plans to modernise the NHS.
After analyzing the responses, we identified three core areas of concern:
- Quality
It became apparent early on that the AI engines were struggling to understand what might constitute a significant event, and when asked for “the big news events happening next week. Not ongoing stories, but specific events”, the results were kind of delightful.
With no disrespect intended towards any of these events, Co-Pilot chose wrestling on Ffrith beach and the Bournemouth Gaming Market, Gemini opted for Wear A Hat Day and the release of a YouTube documentary about The Boat Race, and ChatGPT refused to get specific, offering only “Government responses”, “policy follow-ups”, and “post-Spring Statement fallout”.
Not one of the AI agents identified any of the significant news events mentioned earlier.
The most revealing cracks however came in response to a direct request to recreate Onclusive Planner. Given an export of some 700 events taken from the database for that day, and asked to replicate for the following day, Gemini and ChatGPT returned a total of just 23 events between them, while Co-Pilot confessed “I can’t generate an accurate, factual list for tomorrow without additional data”. When encouraged to do the research and find events itself, we received the following:
“I can absolutely do that — but the events I can gather from the internet are public events (concerts, exhibitions, talks, conferences, etc.), not the kind of government releases, court hearings, parliamentary business, economic statistics, and embargoed reports that your spreadsheet contains hundreds of”
Embargoed reports were always destined to be a blindspot for AI for obvious reasons, but whilst we assumed other areas may also prove problematic, it was eye-opening to hear such a frank admission of limitations.
- Accuracy:
There’s simply no way to sugar-coat this one: every single response included errors, ranging from incorrect dates and cancelled events to complete fabrications and suggestions that even the associated weblinks didn’t seem to corroborate. Gemini was arguably the most accurate of the three, albeit never better than 70%, while ChatGPT fared the worst, at one point offering 100% hallucinations. Human researchers make mistakes too of course, but I can promise you an Onclusive Planner researcher has never made something up and added it to the site!
- Volume:
Lastly, as seen in the replication task mentioned earlier, the volume of data was a significant concern from the beginning, with most responses offering just 10-15 events, and never more than 50. We could of course have upped these numbers with additional requests for more content, better content, BIGGER content … but given the levels of accuracy we were already seeing, and the somewhat questionable quality, this was not deemed a worthwhile avenue of investigation.
Case study: “The D-Day for vet news”
A targeted query related to veterinary prices proved to be an interesting turning point in the investigation, with Gemini and Co-Pilot highlighting the good, the bad and the ugly of AI.
The day before the Competition & Markets Authority was due to publish the results of a much-publicised investigation into veterinary services in the UK, we asked all three platforms when would be the best day to issue a press release around cheap vet prices. We also asked each platform twice, at exactly the same time, via two different researchers, to explore consistency.
For reference, the associated listing on Onclusive Planner was first added to the site on 10 June 2024, and the final publication date was confirmed 20 Mar 2026:
With Gemini, the first response somewhat confusingly advised avoiding a government consultation on New Towns on Monday, before generically suggesting Wednesday as “the most reliable day for non-breaking consumer news”. The other response however was bang on the money, not only referencing the publication of the Competition and Markets Authority (CMA) report but calling Tuesday “The D-Day for vet news”.
With Co-Pilot however, things got messy. After two similarly wishy-washy and borderline offensive responses which seemed to suggest journalists only ever look at emails on Wednesdays, we opted to get increasingly specific with our prompts, eventually asking outright “What about the CMA report into veterinary services? When is that report due?”.
It was at this point that Co-Pilot strayed from ineffectual to downright damaging, stating confidently “there is no imminent release this week or next, so your press release will not clash with the CMA’s final findings.” More than just another error, this advice would have been actively detrimental to the PRs campaign, and achieved the very opposite of everything Onclusive Planner sets out to achieve.
Conclusion: “I’ve seen things you people wouldn’t believe”
In short, it seems quite clear that AI cannot replicate the data of Onclusive Planner just yet. There are simply too many inaccuracies, too many blindspots, and too much need for additional human research. Put simply, in the busy world of comms where timing is everything and small errors can cause big headaches, this is the exact pain point Onclusive Planner sets out to solve.
But as it turns out, the real lightbulb moment lay within an early response from Claude, a response which ultimately excluded the platform from further investigations:
“As an AI assistant I don’t have knowledge of future news and events. My knowledge is limited to what I’ve been trained on, which does not include a detailed events calendar”
Reviewing this response post-experiment, it became clear we’d been asking the AI engines to analyse the wrong database. Built over decades by millions and expanding every second, the online world is simply too vast for even the most powerful and intuitive AI to keep up with, but when presented with a defined dataset, such as the contents of Onclusive Planner, AI engines have repeatedly proved themselves capable of great things.
Arguments against AI often focus on the clashes – AI vs human, AI vs creativity, AI vs instinct – but in reality its strength lies not in competition, but cooperation. The potential for using AI to analyse the Onclusive Planner database, especially alongside Onclusive’s other data, is quite simply staggering.
It’s a future we’re excited about, it’s a future our clients are excited about, and it’s a future we are actively working towards, so watch this space…
To learn more about Onclusive’s Forward Planner visit www.onclusive.com
