Agent-type AI: 您的个人研究助理

Agentic AI:Your personal Agentic AI: Your personal Three years on since the launch of consumer-facing LargeLanguage Models (LLMs), and the AI conversation has travelledfrom amazement to business function. The initial launchwas thrilling; the future is still tantalizingly close with new So, where is AI already making an impact? What do insightteams need to know to stay ahead without gettingoverwhelmed? And what happens when AI advances to thepoint of working alongside you as a digital coworker? This What’s AI doing for market research? AI, at its core, refers to machines performing tasksthat typically require human intelligence, such as In market research, we’re mostly talking about: —Machine learning: Tools that learn from data andget better at predictions or classifications. —Natural language processing (NLP): Teachingmachines to understand and respond to human —Generative AI/LLMs: Tools that create content(like summaries, follow-ups, or reports) based on These aren’t replacing researchers; they’re enhancingwhat you already do with speed, scale, and a few The AI explosion What are AI agents? Within the wider world, agent-based tools and systemshave been around for quite a long time. Everyone has hada frustrating experience with a poorly designed customerservice chatbot, looping options, or being hung up on 2024 was the year that AI seriously took root in MR.From conferences to client briefs, it was the topic thateveryone was either excited about, overwhelmed by, or Now, when conferences are asking speakers to chillout on the AI topics, we can see where it’s sticking and —Customer-facing AI chatbots are revolutionizingcustomer service, despite how you might feel about —A growing appetite for automation to removerepetitive manual drudgery. Several AI tools cansummarize meetings and list —More teams adding AI into different stages ofresearch permanently (like survey analysis or —Scheduling tools can help manage your workcalendarand sync up teams across large —A big shift in tone: Less hype, more practicalexploration, asking what AI can do for research, not —Personal assistants can manage your shopping list We already know that AI is here to stay. You may loveor hate some of these changes, but you can’t denythat they’ve had an impact. Insight teams that lean in AI assistants are becoming a mainstay of workflows.Copilot, for example, is styled as a personal assistant thoughtfully and strategicallywill be the ones that thrive. you can have a natural conversation with. However,it can’t yet send an email, book anything, or do much but struggle with Excel, so PA jobs are safe for now. It’sgoing to take a bit of a leap before the tech can catchup with everything we need it to do, and the same goes The idea behind agentic AI (sometimes also referred toas AI agents or AI assistants) is that they’reautonomous systems designed to perform complex tasks from acollection of capabilities in the order they deem fit, with But, despite sharing a similar name, these aren’tagentic AI. These are the building blocks for agentic, Agentic AI leverages large language models and otherAI/machine learning models to interact with users,systems, andother agents. This network of informationmeans agents can use complex reasoning skills, What is agentic AI? The phrase ‘agentic AI’ has started popping up inrecent months, but what does this mean? As the latest buzzword, it’s a bit of a wild west withevery new capability being labelled as ‘agentic’, whenit’s nothing more than a generative model dressed up Two key terms here are “orchestration” and“choreography”, which, combined, give agentic autonomy: The difference with agentic is autonomy and flexibility.In traditional agents, or software systems, the orderin which instructions are carried out follows a fixed,predetermined path. There is no variability. Even if you Orchestration: The central point of contact andcontroller for the system. Requests are routedthrough this into the relevant tools. E.g. the function the right prescription without costing you your will tolive while you spend three hours on hold, it might justsave the patient experience and become humanity’sgreatest invention. So, apply this concept to research,and you have an assistant in your pocket who can Choreography: A sense of freedom. The AI will makedecisions preemptively based on current system The LLM innovation that made this possible is theunderstanding of context and memory. And a lotof really, really smart coding and modelling. Thinkabout your ChatGPT, Gemini, Claude or other LLMexperiences. You can follow up, say ‘more’ or ‘notquite like that’ without re-entering the whole prompt. Traditional to generative to agentic Traditional AI: Single-task specialists.Think spam filters, predictive text, or This is what stops you from having to repeat yourselfso much, put information in a certain order, or getfrustratingly stuck in a loop of “sorry, I didn’t quite Generative AI: Conten