Rutherford Soddy

Ethan Mollick consistently writes some of the most insightful perspectives on generative AI. This blog post mirrors a lot of what we’re learning in working with our customers – the best ideas for how to apply AI come out of our collaboration with on the ground experts. Great read for any companies interested in building internal GenAI capabilities. Latent Expertise: Everyone is in R&D - by Ethan Mollick (oneusefulthing.org)

Does an AI chat application actually make you more productive? Since ChatGPT, we’ve all seen too many chat applications. They often look like the OpenAI playground. They might include a feature to upload and search through documents. Sometimes they offer other integrations, like prompt libraries where users can store prompt templates and prompt chains. One of the reasons for the similarity of these apps is that they’re not hard to make. OpenAI has user-tested GUIs that are easy to mimic. Open-Source packages like Streamlit, LlamaIndex, and LangChain—all fantastic tools—make it effortless to design an interface, build vector databases from uploaded documents, and connect to a language model. And users have been primed to want these apps. But is this the best way to use AI in your business? Some tasks are certainly easier with a copilot: rewrite the below paragraph, summarize this document, find me an example of, give me ideas for. Anyone who uses ChatGPT for these things is getting a nice productivity boost. However, for more complex tasks, a chat interface can be counterproductive: ▪ It takes a lot of time to manually push content through a series of steps ▪ You might need to pull in data that can’t be uploaded as a document ▪ You usually can’t specify different search methods across documents (there is a big difference between searching for a small fact, like a product number, and a complex concept, like a technical approach) ▪ Your individual work usually isn’t scalable across teams In some sense, a chat interface shifts all the work of developing a workflow to a user who may not be an expert prompt engineer, and who probably doesn’t understand all of the possibilities for customizing and evaluating document search methods, leveraging different models for different sub-tasks, automating external data connections, and integrating the LLM with other application features. At Rutherford Soddy, we don’t put this on the end-user. We’re not deploying a tool that can do “anything.” We make tools that actually do something. We work with our customers to identify a targeted task, analyze the steps needed to complete it, identify slices of the workflow that AI can handle, and build programs to do the work at the push of a button. The goal is to give employees “Accelerators” to help them throughout the day. Generative AI is still a new technology. It’s not perfect, and it can’t do everything. But we can use it to solve more abstract problems than ever before. And that helps teams work faster, improve quality, and spend more time on their hardest problems. For another take on this topic, we recommend this fantastic LangChain post on non-Chat applications for AI (https://lnkd.in/dUmaucv7).

"More proposals fail because of poorly executed or nonexistent Pink Teams than any other cause." #ThinkPlanWrite

Some really exciting advances in Generative AI have come out this fall. The pace of innovation is accelerating. All businesses should take note. Problems that were difficult six months ago are effortless today, and application features that once seemed impossible are now on the horizon. Many of these innovations have only been visible in the developer communities, but we expect the resulting applications to impact knowledge workers at businesses of all sizes. To help sort this all out, we’re posting a quick summary to highlight four of the major developments (including new tools and platforms) we’ve been tracking and using. Some of them are more technical, but these are concepts that all business leaders need to start learning about. Feel free to reach out to us to discuss business impacts and possibilities.

This is a game changer. Many similar applications already exist, but OpenAI is making it much easier to develop custom AI assistants. The potential use cases for business workflows are endless. https://lnkd.in/gjxKzFfg

To show the value of PB&J, Rutherford Soddy is going to start offering free samples of solution reports to interested contractors on some major acquisitions. Be on the lookout for announcements in the next few weeks, and follow us to make sure you don’t miss out! And in the meantime, for anyone who hasn’t tried out PB&J yet, we made the below slideshow to explain what it is. Email support@rutherfordsoddy.ai to request a discussion.

What makes a proposal successful? To find out, we ran pilots with some of the highest performing capture and proposal professionals in the US Federal Contracting industry. We studied how they work, and we identified some common themes. First, for their best proposals, they use a “Think, Plan, Write” method: a structured process to iteratively develop and review content. Often, the first draft of a proposal is just bullet points describing the approaches to each section, sometimes called a “Pink Team” or a storyboard. This allows internal reviewers to strategize on solution components, align customer messaging, and course correct on any issues before the pages are written. However, even the highest performing organizations admit that reality sets in far too often. Their team gets behind the 8 ball for various reasons, and the chaotic scramble that we call “working a proposal” starts. We’ve all been there: the upfront thinking and planning gets rushed or skipped entirely. Sections are written, thrown out, and hastily re-written. The stress builds. And the proposal winds up getting finished by a few people trying to save the day. These proposals almost always lose, and everyone is threatening to quit by the end. We all know there’s a better way, and technology can help. With Winning in mind (and less chaotic/stressful proposals as a nice side benefit), we developed our first product, PB&J (Proposal Brainstorm and Jumpstart). It delivers an initial cut of solution approaches, customer benefits, and more detailed technical features within a few hours. This gives your team a rapid starting point so they can create a plan of attack for each section, and focus their time on strategically improving the solution. And that means more winning proposals. Email us at support@rutherfordsoddy.ai to request more information

Rutherford Soddy is excited to announce PB&J (Proposal Brainstorm & Jumpstart), an AI-powered application that accelerates Business Development and Proposal Teams! Think of PB&J as a customized research report for your RFP, with solution ideas for your storyboards or pink team drafts. Delivered overnight. And for less than a few hours of B&P cost. One or two fresh ideas can be the difference between winning and losing, so the ROI is huge. Visit our website at https://lnkd.in/ewcYD7eu to check out a sample and learn more about how PB&J can accelerate your proposal teams! Get off the blank page quickly—Just order up a PB&J! #PB&J #GetOffTheBlankPage #AIforBusinessDevelopment #GovCon #Proposals #Pwin