Getting more out of the AI tools you already pay for

Adapted from the update iNDustry Labs sends the AI Executive Group, a group of manufacturing and logistics leaders across the South Bend–Elkhart region.

Key points from the Q1 AI event

• AI can boost individual staff productivity and radically improve a specific business function, customer service or procurement for instance. Both are worth pursuing.
• Do not spend months building a business case for enterprise LLM licenses. LLMs are what Excel was 20 or 30 years ago, and most companies never evaluated the financial return they were getting from Excel.
• The SaaS-pocalypse is already here: a regional company rebuilt its project management tool in 60 hours for a fraction of the subscription cost.
• As programming gets further abstracted, new roles around AI architecture and orchestration are emerging.
• 60% of today's jobs did not exist in 1940. New roles such as AI product manager are evolving for the same reason, an effect known as Jevons paradox.
• The first software cost estimate is rarely the final one. Consider different approaches before killing an idea on cost: several simpler orchestrated agents and one complex agent can differ enormously in price.

Getting the most out of the assistants

Most companies in the group are already paying for at least one AI assistant and using a fraction of it. What the features actually do:

Claude

• Chat. The starting point. Be specific about what you need: "summarize this report for a plant manager deciding on a second shift" beats "summarize this report." Upload files directly, since PDFs, spreadsheets, images and Word documents are handled natively. Do this under a paid company license to keep your data private.
• Artifacts. When you ask for something built, a chart, calculator, formatted document or interactive tool, it renders as a live product in the conversation so you can iterate without starting over.
• Projects. Group conversations, upload reference files and set persistent instructions for a recurring engagement. Every new chat inherits that context. This is what most of the iNDustry Labs staff use inside client environments, so context stays with the business after a project closes.
• Skills. Reusable instruction sets that encode your team's processes. We built skills for drafting project plans, applying brand guidelines and running technical analyses. Once built they fire automatically when a relevant request comes up.
• Connectors. Link the assistant to Microsoft, Slack, Gmail, Calendar, Airtable and more, so you can say "find the Q3 deck in my Drive and summarize the key metrics" instead of hunting for slides.
• Cowork. Assign a full task, "organize these receipts into a spreadsheet," "compile a competitive analysis from these five documents," and it plans, executes and delivers. It reaches your local files, connected apps and browser, and tasks can be assigned from a phone. Useful where connectors lack native access and more human-like actions are needed.
• Claude Code. A command-line tool with full access to your codebase for writing, debugging, refactoring and testing. Built for developers, though plenty of non-technical people pick it up quickly.

ChatGPT equivalents

• Chat and Canvas. Similar upload and chat experience, with Canvas adding a co-editing workspace for documents and code, roughly analogous to artifacts but with more of a document-editor feel.
• Projects. Available across plan types, with shared projects on Business and Enterprise.
• Custom GPTs. The equivalent of skills: a specialized assistant with custom instructions, knowledge files and API connections.
• Apps. A unified directory connecting to Google Drive, SharePoint, Notion, Linear, Slack, Jira, Salesforce and others, with synced connectors auto-indexing your data.
• ChatGPT Agent. Web browsing and deep research merged into one agent that browses, runs code, analyzes data and produces decks and spreadsheets from a single request. Cloud-based rather than desktop-based.
• Codex. The coding agent, available as a desktop app, CLI, IDE extension and web interface, supporting parallel agents on different branches of the same repository.

How the platforms compared

A snapshot of the three major assistants as they stood in April 2026. Model versions move quickly, so treat the bottom rows as a point in time rather than a current spec sheet.

Shared by group members

For developers: knowledge graphs

As your insight needs advance, it is worth thinking about ontology within your systems rather than spending all your time writing complex queries.

For engineers: where Autodesk is taking AI

Our team met with Autodesk to hear how they are implementing AI across their products.

• Forma is the AI flagship, expanding into a full architecture and construction industry cloud with neural CAD models that generate editable layouts from simple inputs.
• Third-party AI plus CAD is booming. Assistants are connecting to Autodesk products via MCP servers, enabling natural language control of Fusion 360 and AutoCAD.
• Autodesk Assistant has gone agentic, converting text prompts into editable 3D geometry in Fusion and executing commands in natural language, rolling out across Revit, AutoCAD, Civil 3D and Forma.
• More than 1,600 authorized training centers in the US deliver certified instruction, and state workforce funds may subsidize employer-sponsored training and certifications.
• A generative design for manufacturing certification is available as an advanced credential in Fusion, with a free prep course. A strong upskilling path for manufacturers exploring AI-driven design.