Welcome back to AI, Actually! This week Jim Johnson hosts Andy Sweet, Ada Gil, and Nicole Kosky for a conversation that keeps coming back to one persistent question: where is the ROI in AI actually coming from? The team argues that the answer requires a reframe. Companies aren’t buying technology, they’re buying intelligence, and that […]
AI, Actually – Episode 25: Why AI Agent Pricing Should Be Value-Driven Read More »
Every custom agent harness makes assumptions about what the model can and can’t do on its own. Those assumptions go stale as the model improves. An earlier version of Claude would wrap up tasks prematurely as its context window filled, a behavior sometimes called context anxiety. Teams built context resets into their harness to compensate.
Inside Claude Managed Agents: Architecture, Primitives, and Cost Tradeoffs Read More »
A finance team recently automated its monthly close process after losing its controller. A sales team built an agent that takes a company name and a contact, researches the account, and generates a ready-to-send outreach asset. A marketing team cut the cost of building a client-ready slide deck from $30 to $50 down to under
Claude Managed Agents: What They Are & When to Use Them Read More »
Welcome to Episode 24 of AI, Actually! This week Jim Johnson hosts a deep dive into one of the least understood parts of enterprise AI: what it actually costs to run it. He’s joined by Shanti Greene and Stew Chisam, along with first-time guest Jake Barger, for a wide-ranging conversation on tokens, routing, and the
Most companies automate the month-end close by buying software and hoping the team adopts it. We took the opposite approach: we built the automation around our own close first, on our own books, before we ever offered the Month-End Close solution to a client. Our controller gave notice. She left behind 31 hand-off recordings and
How AnswerRocket Cut Its Own Close Process by 80% Read More »
TL;DR: Every AI agent design starts with the same question: how should this actually be built? For a while, the answer was simple: loops, full stop. The Ralph Wiggum loop, a script that reruns an AI coding agent against the same task until it passes a test suite, became shorthand for that era. It removed
AI Agents vs. Agentic Workflows: When to Use a Loop and When You Need a Graph Read More »
Welcome to Episode 23 of AI, Actually! Pete Reilly hosts this episode, joined by Shanti Greene, Mike Finley, and Stew Chisam for a conversation sparked by a single viral post online: are we still building AI systems around loops, or has the industry moved on to graphs? The team unpacks what that distinction actually means,
Welcome to Episode 22 of AI, Actually. Jim Johnson is back in the host seat, joined by regulars Nicole Kosky, Shanti Greene, and Andy Sweet for a conversation that pulls together three topics that look separate on the surface but turn out to be closely connected: early impressions of Fable 5, the rising scrutiny around
Welcome to Episode 21 of AI, Actually. This week Pete Reilly hosts a conversation with Shanti Greene, Mike Finley, and first-time guest Ada Gil, who leads the AI business transformation practice at AnswerRocket. The discussion is built around a recent video from Nate B. Jones called The Trillion Dollar Agentic Workflow Opportunity Is Here. The
Welcome to Episode 20 of AI, Actually! This milestone episode brings together Jim Johnson and Andy Sweet alongside returning guest Nicole Kosky and a new voice to the podcast, Ben Titmus, who leads the data, platforms and infrastructure practice at AnswerRocket. The topic? The one thing that underpins all of AI: data. The conversation cuts
TL;DR: Stop trying to find the one best AI coding model. There isn’t one. Use Claude Opus or Codex for the hard stuff (architecture, debugging, security), then hand off smaller tasks to cheaper models like Kimi K2.6 or Qwen3 Coder Plus. Budget about $200/month for a premium subscription, pick one tool, and learn it cold.
TL;DR: The forward deployed engineer concept Palantir popularized is being misapplied. AI has shifted the bottleneck from engineering capacity to problem clarity. What enterprises actually need on the front lines is a forward deployed consultant: business-first, AI-fluent, and capable of building prototypes that solve the right problem before any production code gets written. The forward
You Don’t Need a Forward Deployed Engineer. You Need a Forward Deployed Consultant. Read More »
Welcome to Episode 19 of AI, Actually! This week Jim Johnson hosts Shanti Greene, Nicole Kosky, and Stew Chisam for a conversation about one of the more talked-about terms in enterprise AI right now: the forward deployed engineer. The team sets out to cut through the noise and get at what this role actually means
Welcome to Episode 18 of AI, Actually! This week, Pete Reilly hosts a candid conversation with Mike Finley, Nicole Kosky, and newcomer Michelle Hamilton, who joins AnswerRocket fresh from years of helping organizations navigate AI change management. The topic: why some companies are stuck in pilot purgatory while individual employees are quietly getting 10x more
AI, Actually – Episode 18: How to 10x Your Individual Productivity with AI Read More »
This is Part 2 of my series on the 7-Layer Agentic AI Framework. Part 1 covered the Perception Layer, how agents ingest and normalize raw inputs. The Semantic Layer picks up right where perception leaves off: perception handles what the agent sees; the semantic layer governs what the agent knows and what it’s allowed to
The Semantic Layer: How AI Agents Understand The Context of Your Business Read More »
This is Part 1 of my series on the 7-Layer Agentic AI Framework. This post covers the Perception Layer, how agents transform raw inputs into structured intelligence before any reasoning happens. Part 2 covers the Semantic Layer, which picks up right where perception leaves off: perception handles what the agent sees; the semantic layer governs
The Perception Layer: Why Your AI Agent Needs to See Before It Can Think Read More »
Welcome to Episode 17 of AI, Actually! This week features Jim Johnson as host, joined by Andy Sweet, Nicole Kosky, and first-time guest Evan Gatch—AnswerRocket’s VP, Consulting Sales—for a candid conversation about what consulting actually looks like in the AI era. The spark? A recent Wall Street Journal piece validating what the team has been
New Researcher + AI brand tracking methodology is the first of multiple initiatives designed to drive deeper insights and improved outcomes for clients. Highlights: LONDON AND ATLANTA, March 18, 2026 – Basis Global, a market research and brand intelligence firm, today announced a strategic partnership with AnswerRocket, an enterprise AI solutions consultancy, to redesign how
Welcome to Episode 16 of AI, Actually! This week Pete Reilly hosts alongside Mike Finley, Shanti Greene, and Stew Chisam to dig into one of the hottest debates in tech right now: is vibe coding the death of software development, or just the next chapter? The headlines have been dramatic—Block, parent company of Square, announcing
AI, Actually – Episode 16: Will Vibe Coding Kill Software Engineering? Read More »
Welcome to Episode 15 of AI, Actually! This week Pete Reilly hosts Andy Sweet, Jim Johnson, and Stew Chisam for an honest look at the growing gap between what’s possible with AI today and where most businesses actually stand. While the team is building full-stack apps in an afternoon and watching agents autonomously run multi-hour
AI, Actually – Episode 15: How Businesses Can Actually Get Started with AI Read More »
Complexity is crushing the CPG industry. Across the value chain—from SKU proliferation to global supply logistics—CPG leaders are overwhelmed by raw data but severely lacking decisive, actionable answers. The systems built for the last decade cannot manage the scale of modern business, leaving up to 90% of your operational decisions running blind. Here’s the simplest
How Agentic AI Unlocks Growth and Infinite Analyst Scale for CPGs Read More »
Welcome to Episode 14 of AI, Actually! This week Pete Reilly hosts Jim Johnson, Shanti Greene, and Stew Chisam for a conversation about three interconnected developments forcing enterprises to rethink everything from organizational structure to software procurement. The discussion opens with OpenClaw (formerly ClaudeBot and briefly MoltBot)—an open-source AI assistant that doesn’t just suggest actions
Welcome to Episode 13 of AI, Actually! This week takes a different format—Pete Reilly hosts a live demo and discussion with Alon Goren, Mike Finley, and Andy Sweet as they walk through a working CRM built in just a few weeks using AI coding agents. But this isn’t about showing off a new CRM—it’s about
Featuring Vaithi Bharath, Associate Director of Data Science & AI Solutions at Bayer. In a recent episode of the AI in Business Podcast sponsored by AnswerRocket, Vaithi Bharath, Associate Director of Data Science & AI Solutions at Bayer, joined host Matthew DeMello, Editorial Director at Emerj AI Research, to explore one of pharma’s most persistent
Welcome to Episode 12 of AI, Actually. This week features Jim Johnson as host, joined by Joey Gaspierik, Nicole Kosky, and Stew Chisam for a deep dive into what might be the most important emerging discipline in enterprise AI: Agent Operations (AgentOps). As enterprises move from impressive demos to production agents doing real work, a
Co-authored by Stew Chisam, Operating Partner, StellarIQ If you blinked recently, you might have missed a major shift in the AI landscape. In the span of just a few weeks, the industry has delivered a rapid succession of sophisticated releases: OpenAI’s GPT 5.2, Anthropic’s coding specialist Claude Opus 4.5, and Google’s efficiency engine, Gemini 3 Flash.
Beyond Benchmarks: How to Choose Between Gemini 3, GPT 5.2, and Opus 4.5 Read More »
Welcome to Episode 11 of AI, Actually! This week features Pete Reilly as host, joined by Jim Johnson, Alon Goren, and Shanti Greene to unpack OpenAI’s recent white paper “From Experiments to Deployments: A Practical Path to Scaling AI.” But this isn’t just a review—it’s a reality check based on years of front-line experience helping
The AI landscape is evolving faster than most enterprises can adapt. We asked our leadership team for their predictions on where AI is heading in 2026. Here’s what they had to say. 1. Multimedia Creation Will Flawlessly Mimic Real Artists and People Video, music, and image generation will take another massive leap forward in 2026.
Why Karpathy’s “Decade of the Agent” Means You Need to Start Today Andrej Karpathy just handed executives everywhere what might sound like permission to wait. In a recent Dwarkesh Patel podcast, the AI expert said 2025 isn’t the “year of the agent,” it’s the “decade of the agent.” I can already hear the boardroom conversations:
Don’t Wait for Perfect AI: Why the ‘Decade of the Agent’ Means Start Today Read More »
Welcome to Episode 10 of AI, Actually! This week features Pete Reilly as host, joined by our technical dream team: Andy Sweet, Shanti Greene, and Stew Chisam. Fresh off Gemini 3’s launch, the team goes deep on what’s actually different, how it stacks up against GPT-5.1 and Claude, and why Google’s play is about ecosystems,
AI, Actually – Episode 10: Gemini 3 Deep Dive and Bold Predictions for 2026 Read More »
