Rethinking work with AI

If “AI” has started to feel like background noise on your LinkedIn feed, Dave (joined by Frank) tries something different here: a practical, no-jargon walkthrough of where to actually start. Dave isn’t a computer scientist — he says so upfront — and instead explains agentic AI with the simplest comparison he’s got: it’s like hiring a new employee, except one that can absorb your processes in minutes instead of months. He draws a clear line between Genius Cortex, the AI built into the ERP that analyzes your existing data and surfaces insights, and agents, a separate layer that can take real actions across systems — reading an email, gathering data from a supplier portal, or spotting an anomaly in a spreadsheet. Rather than pitching a six-figure AI transformation, Dave walks through Darwin, a small tool Genius built internally for its own customer success team, that took quoting time from 15 hours a week down to 5 — built in about three weeks for roughly a $5,000 investment. Throughout, the message stays consistent: AI is bad at judgment calls, so humans stay in charge of exceptions, critical decisions, and strategic direction. If you’ve been wondering where to start with AI without committing to a full process audit, here’s what was covered.

What You’ll Learn In This Webinar

Chapter 1 — Real-World Applications of AI

Dave sets the agenda: define agentic AI, position it against Genius Cortex (the AI already embedded in the ERP), talk about the role of humans, then walk through a real internal use case before opening the floor for questions.

Chapter 2 — The Role of Humans in AI

Dave previews that human fear around AI will be addressed directly, then introduces the core analogy for the rest of the session: agentic AI is best understood as another employee — one you still have to train, except its ramp-up happens in minutes rather than the 3-6 months a new hire typically needs, observing incoming data and making decisions based on the guardrails it’s given.

Chapter 3 — Genius Cortex Overview

Dave distinguishes embedded AI (Genius Cortex) from agentic AI as a deliberate positioning decision at Genius. Cortex’s first release helps with system usage questions; future versions will do internal data analysis and surface patterns (like a vendor that’s consistently late) — but all of this depends on data actually being entered into the system, which Dave admits doesn’t always happen given time constraints.

Chapter 4 — The Functionality of AI Agents

Agents act like fast-learning employees that gather external information (mailboxes, portals, spreadsheets, even IoT data from shop-floor machines) and take action across multiple systems, since no single ERP will ever be a business’s only system. Dave notes this agentic layer is needed because the ERP stays fairly generic, while each business has its own tailored processes spanning multiple departments and systems.

Chapter 5 — Human Supervision in AI Processes

AI is genuinely bad at judgment calls because it never has the full context an owner has about growth goals and investment priorities. Dave gives an example of a recommended reorder quantity (25 units instead of 5 based on seasonality) that still needs a human sign-off given the cost impact — a decision he notes an AI shouldn’t make alone.

“AI is there to augment humans, it’s not there to replace them.” — Dave

Chapter 6 — Conversational AI Features

Cortex is moving toward a conversational interface — asking it directly to filter orders or show a customer’s history — and toward trend detection, like flagging that a customer’s orders dropped 15% this quarter. Dave previews that Version 18 will add an AI-driven recommended reorder tool that suggests delivery dates based on historical patterns; Frank adds that future versions may also surface which shop employees are performing best, as a model for others.

Chapter 7 — Data Management with AI Agents

The agent layer is especially good at handling unstructured data — reading an email or a sales order and pulling out what’s actionable, including legal clauses like late-delivery penalties. Dave describes a project built for a customer whose team was manually managing a weekly spreadsheet of order date changes: the tool applies valid changes automatically and flags anomalies (like a 45-day date shift that doesn’t make sense) for a human to review. Frank estimates that roughly 70-80% of this kind of process can be handled this way, with the remainder needing human judgment.

Chapter 8 — Human Touchpoints in AI Operations

Dave narrows the human’s role to three touchpoints: handling exceptions the AI flags, validating critical or strategic decisions (like approving a major equipment purchase), and providing context AI will never fully have. He notes AI adoption remains slow largely due to fear of replacement, and reframes AI as something meant to elevate people’s work rather than replace them.

Chapter 9 — Open Discussion on AI Implementation

Frank reminds attendees to use the Q&A, and Dave pivots into Genius’s own internal AI journey — a strategic review started about a year earlier to figure out where to start applying AI to their own business, which leads into the Darwin example.

Chapter 10 — Introducing Darwin: AI for Customer Success

Genius’s customer success team fields about 300 tickets a week, roughly 25% of which are simple quote requests; a maternity leave and medical leaves temporarily shrank the team from five to two people, exposing that quoting alone was taking about 15 hours a week. Rather than building a tool to read and act on an entire email automatically, the team chose the smallest viable step: Dave jots down the requested items from an email (like 5 full licenses and a Shop Floor trial) into a quick note, sends it to Darwin to generate the quote, and spends the freed-up time replying personally to the customer to maintain the relationship — with the finished quote auto-attached to the HubSpot ticket, ready to convert to an order.

Chapter 11 — Efficiency Gains with AI Solutions

Quoting time dropped from about 15 hours a week to about 5, freeing roughly 10 hours for strategic reviews, growth conversations, and relationship-building rather than repetitive data entry. Dave uses this to illustrate that meaningful AI projects don’t require six-figure investments — Darwin itself cost roughly $5,000 to build.

Chapter 12 — The Impact of AI on Job Markets

Dave points to a chart showing S&P 500 growth historically tracking job openings — until ChatGPT’s November 2022 launch, after which that link decoupled: company growth continued without a matching rise in hiring. His suggestion: rather than auditing an entire shop for AI opportunities, look first at any role you’re about to hire for, since that’s already a proven pain point — a $5,000-15,000 AI project can sometimes replace the need for a $50,000-125,000 hire. Frank adds that narrowing scope (like Darwin’s “baby step” approach) delivers large benefits without over-engineering the first project.