Alastair Gill

One of the hot topics on the lips of those close to the AI space is Agent AI or, as some in the industry are calling it, Agentic AI. For me, Agent based AI systems are a hugely exciting opportunity to really scale how we apply AI within organisations in a way that is cost and time effective, manageable and gives us the opportunity to truly optimise businesses processes through the embedding of various forms of AI agents to solve specific elements of complex tasks. In this model we are not just thinking of converstional AI's but AI's of many different types including ML, Optimisation, Automation, Monitoring, Security and other functional agents. Many of the frameworks available on the market are solely considering the big foundational models alongside small language models but in reality the value comes when those models are able to trigger other types of agents to execute a specific function and for them to process the response. The challenges we will face in this new Agent AI landscape will include, as a primary element, then need to design a consistent communication framework so that agents developed in one platform or technology, will be able to integrate and "talk to" agents developed elsewhere, as well as an effective monitoring mechanism across the ecosystem deployed to identify any agents not performing as expected and trigger alerts to take relevant actions (actions by other agents or humans). It's a fast evolving space and I'd be excited to hear what others think or are doing here too. Please share your thoughts! #ai, #aiagents, #agenticai, #innovation, #avanade,. #datascience, #emtech, #microsoftai

56

4 Comments

We've just launched a major new feature Ballpark called Data Tables - a powerful way browse, analyse and export your research responses. https://lnkd.in/eCf3Uwxw Why spend time building a table?! • As our customers grew, so did the volume of responses gathered for surveys and usability tests, going from tens of responses, to hundreds and now thousands. The only way to parse this effectively was to export to a spreadsheet and do your work there, which sucked. We added the most used functionality of a spreadsheet to Data Tables. • Customisation of the data is super important - everyone has different ways they want to view information. Data Tables allow you to show/hide columns and export based on the view you've set up. • Performance is supercharged, thousands of responses load within a second Now live on every plan https://ballparkhq.com/ #userresearch

55

5 Comments

We’ve all used Wikipedia at some point to get a quick answers, background research, or just to dive into random topics. It’s a rare gems on the internet that’s still completely free and ad-free, and it’s been built and supported by regular users like us. Only 2% of Wikipedia readers actually donate. That’s how they keep it going without subscriptions or ads. So, if Wikipedia’s ever helped you out, now’s a great time to give back. It’s not just about the everyday personal stuff. As business professionals, we turn to Wikipedia for reliable information when we need it most. Whether we’re researching industries, researching algorithms, learning new terms, or just getting a quick insight of a region or market player, Wikipedia is a go-to tool. It's a huge social resource; accessible to everyone, everywhere. If you want to help keep it this way, please consider donating. Even a small amount goes a long way to ensure that Wikipedia remains free for all. Let’s help keep the world’s knowledge open and accessible for everyone. 🙌 #SupportWikipedia #FreeKnowledge #GiveBack #DonateNow

3

1 Comment

**Using AI to Catch Littering from Vehicles...** **LitterCam** uses AI to detect litter as small as a cigarette butt. The UK employs this tech to fine offenders by matching footage with license plates. Despite privacy concerns, the cleaner environment and accountability benefits make it worthwhile. 👇👇👇 https://lnkd.in/d2WH_uc6 #microsoftpowerbi #dataanalysis #PowerBI #datascience #dataanalyst #GenAI #MachineLearning #AI #artificialintelligence #QuizTime

1

How are you using Large Language Models (LLMs)? We've been testing lots of applications for LLMs, and it’s been fascinating to see their strengths and limits. Brilliant at technical tasks like formulas for data analysis, html code for website development, and writing complex formulas for tableau. But when it comes to creating polished, personal content—like a Social Media post, or anything that needs a human voice—LLMs often need a lot of fine-tuning. That’s where the human touch really makes a difference. Teams can quickly become frustrated and not see the potential if used unsuccessfully. Megan Maya and I will hosting a webinar this Thursday on how charities are adopting AI in their organisations. We'll even share how easy it is to customise a bot like Alex the Analyst. Our biggest win has to be used for internal efficiency and deeper research and data analysis much more than consumer facing communications. Join the webinar 👉 https://lnkd.in/g5rueZ9 Complete our State of AI survey: https://lnkd.in/gzmweH49 #AI #LLM #Automation #DataAnalysis #Webinar

22

3 Comments

Alot of academics have also known this for decades. Research from Emily M. Bender and Melanie Mitchell show the risks of focusing on language 'form' without meaning. Cognitive science innovations in RRG Linguistics and brain science, Patom Theory, combine to link 'semantics to syntax' for implementation on machines. The human brain's ability as a baby to acquire their 'mother tongue' shows the blueprint to capture this law of nature, that is language. Let's use it for the trustworthy *and* last interface we will ever need. #patomtheory #RRGLinguistics

4

LLMs can run twice as fast by removing redundant attention layers that don't impact performance materially. https://lnkd.in/e-sSTqSw ----- Newsletter https://lnkd.in/emCkRuA More story https://lnkd.in/enY7VpM LinkedIn https://lnkd.in/ehrfPYQ6 #AINewsClips #AI #ML #ArtificialIntelligence #MachineLearning

3

✍️ A couple of things I've been working on have been published recently. One is an article for Harvard Data Science Review that I wrote with Elena Simperl and Sir Nigel Shadbolt. It makes the case for drastically increasing transparency around the data used to train foundation AI models. It's admittedly a little out of date, as we submitted back in 2023. But there's still plenty of work to do to ensure we really know what this wave of AI has been trained on. There's more interesting articles in this special edition, 'Future Shock: Grappling with the Generative AI Revolution'. https://lnkd.in/eKapqiBX The other is on intellectual property, with Gefion Thuermer, Ben Snaith, Thomas Carey-Wilson and others at the Open Data Institute. It argues that a fair intellectual property regime is key to creating a data and AI ecosystem that works for everyone. It recommends that an incoming UK Government avoids reaching for a one-size-fits-all solution to the copyright conundrum, supports collective rights management and seeks international alignment. https://lnkd.in/eJBJE8p4 Please share and let me know if you'd like to talk.

52

The next generation of personalised digital services will use data ecosystems and AI to build fairer and more profitable services. This short article on Medium outlines a few thoughts about what they will look like.

8

https://lnkd.in/eSNY4SWn Gen AI is not a sustainable flavour of AI, by so many metrics. But let’s talk about real people mechanical turking the hell out of training the huge swathes of data LLMs require.

5

Check out William P. A. Morrissey session on scaling qualitative insights with GenAI!

3

Been using LLMs as part of my dev workflow for a while now - mainly for the stuff that takes time but doesn't need creative thinking. They're great at writing comprehensive docstrings that actually explain what the code does, with solid examples and edge cases. Been finding them particularly useful when refactoring - once I spot related functions that operate on the same data, they help with the heavy lifting of converting them into proper classes. Makes maintenance a lot more manageable, especially with legacy code. Anyone found other interesting uses? #SoftwareEngineering #Programming

18

3 Comments

Not only does Stardog Designer support creation and full lifecycle maintenance of ontologies via LLM prompting, but the new release also supports ontology modules so enterprise teams can evolve parts of a model independently, promoting reuse and sane semantic integration across large, complex orgs.

51

1 Comment

I didn't think that Generative AI was really going to impact Data Engineering until I met Uriel Knorovich in Verbier con Chris Tabb, Mark van der Heijden, Joe Reis 🤓 etc. and other great folks One of the biggest problems we data folks know about is schema change. Schema changes - pipeline breaks. This problem is notorious in the web scraping world. A HTML document can have a deeply nested structure. If someone changes the HTML doc structure, your webscraping script fails. This is a pain to maintain as the owner of the web page isn't someone you know (normally) - #datacontracts ? Please. It's not generating anything massive *per se*, but with AI you can feed it A) a HTML doc (or the raw data) and then B) a Schema and it is pretty good at accurately returning the correct data i.e. getting around "schema change". Best part is, with the right prompt, you can ensure the format (if not the values) are always the same, which makes the process more robust. Why can't this be applied to data ingestion en masse? Why can't we use Gen AI to solve Schema change? Why can't we feed CDC data into an LLM so it can intelligently work out that the connection_guid column that just got dropped got replaced by connectionid? I am sure there are some good reasons. Let me know in the comments 👇 Also sneak peak of what we are working with at Orchestra and web-scraping 🔗 https://lnkd.in/eBeZxUYa --> go to integrations -> Nimble #dataengineering #dataorchestration

32

5 Comments

💡 The EU AI Act: A Must-Read for Tech Developers Everywhere 💡 For those who haven’t seen it yet, I highly recommend checking out the summary of the EU AI Act here https://lnkd.in/e_mZRa-F. As we continue to develop innovative software in the UK, especially with AI, it’s essential to consider the Act’s implications if your solutions impact EU citizens or the EU market. One specific aspect I’m particularly glad to see is the ban on emotion-recognition applications in the workplace. I despised the idea of an AI interpreting how I’m feeling and reporting it to a meeting host. Knowing this type of AI application was enabled, would have already made me uncomfortable during a meeting. 😒 What are your thoughts on this new era of AI regulation? #AI #EULaw #TechRegulation #ResponsibleAI #Innovation

7

1 Comment

I can't explain enough how fantastic this section was, we learnt about many interesting topics, such as Schemas: How should my data be organised logically? Normalisation: Should my data have a minimum of dependency and redundancy? Views: Which joins will be made most often?  - Non-Materialized  - Materialized Access control: Should all data users have the same level of access?  - GRANT and REVOKE ROLE  - CREATE OR REPLACE ROLE DBMS: How do I choose between all the SQL and noSQL options?  - NoSQL  - SQL and more! The course was very enjoyable and I highly recommend it. #DataScience #DataAnalysis #DBA DataCamp

3

4 Comments

Spent some time today experimenting with GraphRAG - https://lnkd.in/eVqcVr3A It's a really interesting idea. Essentially, by employing a better data structure (like a graph in this case), one can improve the model's semantic understanding of the data. This is because now it has easy access to abstractions it didn't before. I indexed my favourite plasma physics paper that I read thousands of times during grad school and built simple RAGs - one the standard way, and one using GraphRAG and the difference is tremendous! To the question "what are the main themes of this article?", baseline RAG picked out different phrases and gave an answer that on the surface sounds reasonable but isn't quite there. It also mentioned one theme where the model just gets it completely wrong. Whereas, graph RAG did manage to extract a few relevant themes. Caveat: I hacked this together very quickly and haven't spent much time improving the data extraction. I reckon the graph RAG performance will improve a lot with better data. Planning to try and find some time in the coming days to (1) improve the data quality (2) build a nicer UX to easily compare the two applications The code and screenshots of results are on github under MIT license, as always: https://lnkd.in/eVkvJrcM

34

3 Comments

Delving into the dynamics of the Employee Voice in the age of AI has been a deeply engaging and thought-provoking endeavour. As Head of Data and AI at Natter, I focused on developing the methodology for this study, which examines how traditional approaches - surveys, interviews, and focus groups - compare to the emerging potential of conversations. Leveraging advanced natural language processing (NLP), thematic modelling, and LLMs, we transformed 240 discussions between global leaders into structured, actionable insights. These efforts revealed distinct contrasts between methods: + Structured Interviews often identify broad organisational themes, but + Conversations unlock participant-driven perspectives rich in nuance and rooted in psychological safety. A standout technical feature of the study was the use of vector embeddings to align qualitative transcripts with the 96 themes we identified in the discussions. This approach enabled us to extract deeper insights and ensure precision in mapping qualitative data to meaningful outcomes. Key takeaway? Conversational methods are not only scalable but essential for capturing the full richness of human perspectives. They address challenges like survey fatigue and data trust while enabling organisations to hear and act on the true voices of their employees. This study is the first phase of a broader initiative to explore all four listening methods in greater detail, culminating in a comprehensive analysis next year. Collaborating with the Natter team and learning from so many global leaders has been invaluable. For those working in people analytics or data science, this research offers insights into how conversational analytics can transform our understanding of employee sentiment and behaviour. If this resonates with you, let’s connect - I’d be glad to discuss how this approach might be relevant to your work!

30

4 Comments