Intetics’ Post

A key ingredient in successful data and AI planning is building momentum. In our latest e-book, Richard Pugh and others focus on the questions you need to ask to develop an effective data and AI strategy. Download our eBook and get started 👇 https://okt.to/jrad3n #Insurance #Data #Innovation #Endava

BIG ENTERPRET LAUNCH DAY! 🚀🙏 We are thrilled to launch, Wisdom, Enterpret's AI Copilot for customer insights. Wisdom can scour your entire customer feedback dataset—sales calls, support tickets, surveys, social media, reviews, slack messages, Snowflake tables—and convert it into genuinely actionable insights, in seconds. All insights are backed with citations from raw customer quotes, and the actual natural language query setup is inspectable and editable to its core to ensure complete transparency. My favorite workflow with Wisdom while working with our beta customers - Step 1 - Asking Wisdom to plot a trendline of all customer feedback reasons from a particular segment. Step 2—Double-click on a particular trend by asking Wisdom to summarize all the sub-reasons behind one trending issue. Step 3 - Spotting one sub-reason that seems like an immediate fix opportunity and ask Wisdom to write a PRD for that one sub-reason, backed with customer feedback citations for full context. For the last 3 years, we've been focused on converting valuable unstructured customer conversations from all channels into a time series database that is SQL queryable, thus creating an extremely valuable dataset for our customers that's updated continuously. With Wisdom, the power of that data set is unlocked for every one in your team through just natural language! Quick 90-second video to demo how exactly :)

AI blah blah blah Isn’t that all I need to get attention these days? Anyway, I’ve been using Copilot and while it has some really helpful features it seems wrong to have it writing my tests for me. After all, tests aren’t supposed to test that my code does what it does, they are supposed to test that it does what is intended. Copilot has no idea what the intent is of my code. It got me thinking that we really should turn that around. We should write tests and have copilot write the code. After all, if you create tests that verify the business objective then how the code works isn’t of great importance (yeah yeah, performance and maintenance and blah blah blah). I did a quick search on Google on the topic and the very first result was from Allen Helton (link in comments). Seems like I’m not the only one thinking this is a better idea. As Allen pointed out, @mattzcarrey was the person who gave him the idea. Turns out there is an entire industry around this. I haven’t tried this myself just yet, but if I’m going to involve GenAI in my tests this is how I’m going to do it.

Running out of Clay credits way too fast? 🚫 Here are three tips to optimize your usage: 1️⃣ Have a separate table for your contacts and companies If a company has three contacts (rows) in your table, you will be using three times the credits to enrich that company. Instead, enrich companies in a separate table and pull in the data into your contacts table using Clay’s “Lookup Single Row in Other Table” function. 2️⃣ Use conditions You don’t need to run all integrations for a row unless a contact/company meets certain criteria. Set conditions for each of your enrichment columns. Simply use Clay’s AI formula writer to come up with the right script if you’re not familiar with it. 3️⃣ Use Waterfalls in the right order When using Waterfalls, prioritize the use of cheaper integrations or the ones you already have an account with. Bonus tip! Sometimes having your own account and using the API integration with Clay is much cheaper than using Clay credits (Looking at you OpenAI!). What other integrations are cheaper if you purchase directly from the provider?

Updating my book on understanding Germans and their business practices, I'm also striving to get all the details right. The 2019 edition (still) reads: "Germans, when at a restaurant with friends or colleagues, have the habit of splitting the bill in a way that everyone pays for what they ordered (plus the tip). Some people would quickly get nervous if someone were to suggest sharing a bottle of water (which, in Germany, can easily cost up to seven euros ), and rarely would anyone suggest just splitting the bill equally among the number of heads at the table." I consulted Perplexity[.]ai to help me find out the current price level but don't know how to handle the input I got in the references. Perplexity would only give me some general examples of how to quote references and told me that "[s]ince this is unpublished information from an AI assistant without a formal author name, I have used 'Human' as the author name for illustration." How are you handling AI-generated output/information in your references? Or would you skip mentioning the assistant at all when the result confirms what you would have guessed (implicitly known) anyway? In my case, I can quote an article on the topic that Perplexity mentions as a source, so problem solved; still, I am curious to know.

We keep on launching awesome stuff at Slite ! You know the frustration of spending hours manually formatting docs and still ending up with messy, disorganised files? Say goodbye to that headache! 👋   Whether you're working on an important report or just trying to clean up your internal wiki, this smart formatter is an absolute game-changer. Simply upload your document and let the AI do its magic transforming those messy paragraphs into clear, hierarchical sections with bullet points, headings, and better text flow. In mere minutes, you've got a professionally formatted doc that's way easier to digest and use. We can't wait for you to try out the all-new Document Formatter and let us know your thoughts! 🙏 Link in comments!

Another exciting day here Katanemo as we open source some of the "intelligence" behind Arch (https://lnkd.in/gdXpVWEu). Meet Arch-Function, a collection of state-of-the-art (SOTA) LLMs designed for function calling tasks - that match or beat frontier model performance, but offer a ~12x speed improvement and ~44x cost savings over GPT-4 🤯 In simple terms, Arch-Function helps you personalize your LLM application by calling application-specific operations via user prompts. This involves any predefined functions or APIs you want to expose to perform tasks, gather information, or manipulate data. With Arch-Function, you can build "agentic" workflows tailored to domain-specific use cases - from updating insurance claims to creating ad campaigns via prompts. Arch-Function analyzes prompts, extracts critical information from them, engages in lightweight conversations to gather any missing parameters from the user, and makes API calls so that you can focus on writing business logic. Achieving performance on par with GPT-4, these models set a new benchmark in the domain of function-oriented tasks, making them suitable for scenarios where automated API interaction and function execution is crucial. Katanemo Arch-Function collection demonstrates: 🖼️ State-of-the-art performance in function calling 🎯 Accurate parameter identification and suggestion to gather missing inputs 🌎 High generalization across multiple domains and use cases, from API interactions to automated back-end tasks. 📈 Optimized for low-latency, high-throughput performance, making it suitable for real-time, production environments. Learn more about Arch-Function : https://lnkd.in/gY_kKWDN Learn more about Arch: https://lnkd.in/gdXpVWEu **performance measured as set by the Berkeley function calling leaderboard https://lnkd.in/gYRwSxrX cc Taimur Rashid Ammar Hanafi Arjun Chopra Kamlesh Talreja Arslan Ahmad Swami Sivasubramanian Viet Anh Trinh Dong Wang George Mathew Farheen Omar Omar Zia Khan Muazma Zahid Merrin Kurian Sharjeel Sohail Umar Farooq Minhas Anurag Gupta Andrew Ng Clem Delangue 🤗 Ahsen Khaliq Jon Turow George Elissaios Jafar Shameem William Smith Raja Iqbal

With Beautiful.ai's Timeline Smart Slide, easily adjust and reorganize your content for maximum flexibility and efficiency. And with just one click, seamlessly convert it into a Journey Slide—without losing any data!

Tool deflection is a common response I encounter when presenting well crafted, yet often robust and often complex API lifecycle and governance policies and rules. Meaning, people ask for guidance, and I share, they look at it and don't have time to read and learn, and then they ask, have you consider putting it into XXXX tool. Which in about 10% situations is a good recommendation for guidance improvement and delivery to a specific audience, but 90% it is usually just deflecting having to actually learn what is needed to be successful with delivering your API consistently. I see AI being a high level form tool deflection response in this age of sprawling API landscape. People love deflecting the work at hand in hopes tool XX will save them.

Tomorrow a new Event from Mark Wickersham: https://bit.ly/3NxlfHB Future Proof Your Firm on October 22nd-23rd—delivering cutting-edge strategies and the most current insights to help you thrive in this new AI era of accounting. ✅Leverage AI: Most people think AI is a job-killer, but you’ll learn why it’s actually your secret weapon forgetting more done in less time. ✅Confident Pricing: Billing by the hour is killing your profits. Learn how to price based on value and finally charge what you're really worth. ✅Advisory Mastery: How to transform your firm into a trusted advisor and become indispensable to your clients, deliver more value, and secure long-term loyalty ✅Marketing Strategy: Consistently attract high quality clients with a system that works on autopilot. This event is your roadmap to staying relevant, profitable, and ahead in an industry that’s constantly evolving. Grab your ticket now before it’s too late! Don’t miss out on the insights that can transform your practice for the future.