Introducing Drivepoint SmartModel™ Agent, the CPG modeling expert that lives inside your Excel model. Built for consumer brand finance teams. You open the model you already use, click chat, and start working.
It does three things from the first prompt:
- Imports data from your connected sources (Shopify, Amazon, NetSuite, QuickBooks, your 3PL). No CSVs, no copy-paste.
- Rolls your model forward to the next month with one sentence. Tabs advance, references update, formulas hold.
- Answers questions about your model, grounded in the actual cells. "Where am I over budget in Q1?" "What was DTC gross sales last month?" The model becomes the answer, not the analyst.
SmartModel™ Agent comes pre-loaded with 18 skills built specifically for consumer brand finance: variance analyses, TikTok Shop schedules, channel-mix margins, inventory roll-forward. You don't have to teach it what good looks like.
It runs on Claude Opus 4.6 by default. GPT-5, Gemini 3.1 Pro, and other frontier models are one click away.
Your model stays your model. No migration. No new platform. No retraining the team.
AI in finance is working. Here's where it isn't.
Everyone has takes on AI in finance right now. Some overcooked, some fair. What's not in dispute: across the consumer brands we work with, finance teams are seeing big productivity gains from Claude and other frontier models.
Building variance analyses in minutes. Cleaning data exports. Drafting board narratives that used to eat a Saturday. These are real wins. We use the same tools ourselves.
The part that doesn't make it into the LinkedIn posts is what comes after.
Getting AI to work for fun is one thing. Getting it to production quality (every Monday, on the same model, with the same answer, for the whole team) is another. The first 80% is the demo. The last 20% is fine-tuning, infrastructure, error handling, reliability, repeatability. And right now, that 20% is really hard.
The DIY version is almost always a little messy. A little brittle. A little slow. And almost always single-player mode only.
That's fine for where we are in the cycle. It isn't the destination. The destination is a specialized harness: a system built around the LLM that already knows your function, your data, and your team, so you aren't teaching it from scratch every Monday morning.
For consumer brand finance, that's what we just shipped.
Today we're launching the Drivepoint SmartModel™ Agent: your CPG modeling expert, built right into your Excel model.
This isn't a pivot. It's the next step in the work we've been doing for years: building the financial infrastructure that lets consumer brands forecast, plan, and analyze with the rigor of a multi-person FP&A team and the speed of one really good operator.
The agent is the first time the full picture comes together. A frontier LLM. The Drivepoint data layer. And our hard-won understanding of how consumer brands actually run their finances. All of it available where finance work already happens, inside your Excel model.
The last 20% of finance AI
‘Naked’ Claude (and ChatGPT, and Gemini) is a phenomenal copilot for cleaning a CSV, drafting a memo, building a quick model from scratch. Finance teams across our portfolio are using it every day for those things, and they should keep doing it.
Push past these basic use cases, though, and the gaps show up fast.
The model file is too big to paste in. So you re-import every session. The data is stale by Wednesday. The format breaks. You wrote a great prompt last Tuesday and you can't quite reproduce it this Tuesday. You build something good in your file, and now your colleague needs the same answer and you're back at square one. Version control, which was already hard in Excel, gets harder.
And the kicker: only one person on the team knows how to run any of it. The work doesn't compound. It's single-player.
This is the last 20%. Not "AI doesn't work." It works. Just not at production quality, not for the whole team, and not without a real engineering effort to wrap context, data, permissions, and reliability around the model. Most finance leaders don't have time to build that engineering effort, and probably shouldn't.
That engineering effort, for finance, is what we built.
What a specialized harness for finance looks like
A harness isn't a wrapper. It's three things that have to work together.
The data layer. Live connections to Shopify, Amazon, NetSuite, QuickBooks, your 3PL, the source systems your model already runs on. The agent doesn't need you to upload a CSV or schlep your numbers into a chat window. It pulls from the same place your model does.
The skills. A library of pre-built routines for the work that actually happens in CPG finance. Variance analyses. TikTok Shop schedules. Channel-mix margins. Inventory roll-forward. The agent doesn't have to be retaught every Monday morning.
The model layer. A thin, swappable layer for the LLM itself. Use Claude Opus 4.6 today and GPT-5 or Gemini 3.1 Pro tomorrow. Whatever ships next quarter, after that. Plug and play.
The whole thing lives inside Excel via the Drivepoint Add-In. No new platform. No migration. No retraining your team on yet another system. You open the model you already use, click ‘chat,’ and start working.
It's not a new tool to learn. It's your model, supercharged.
What the agent actually does
Three core capabilities that all work from the first prompt.
Import data from any connected source. "Import my latest Shopify sales." The agent pulls the data, writes it to the right tab, tells you what it did. No copy-paste. No re-mapping fields. No API keys. No MCP setup.
Roll the model forward. "Roll my model forward to April." The agent advances every tab, updates references, preserves formula integrity. The work that used to eat your morning becomes a sentence.
Answer questions about the model. "Where am I over budget in Q1?" "What was my DTC gross sales last month?" The agent reads the actual cells, grounds the answer in your model, shows its work. The next time someone Slacks you that question, you don't have to be the answer. The model is.
Two things underneath all of that matter.
Pick your model. Claude Opus 4.6 is the default. GPT-5, Gemini 3.1 Pro, and other frontier models are one click away. If a provider has an outage, you keep working. If a better model ships, you don't wait for us to bless it. Your skills and context aren't locked to one vendor.
It already understands CPG. The agent comes pre-loaded with 18 skills built specifically for CPG finance. You don't have to teach it what good looks like. That's the foundation no generic AI tool has, and it's why this works the moment you click Chat.
What this unlocks
For a 3-person finance team, the agent reclaims hours of analyst time per person per month. Not a new-workflow gain. Removing the workflow that was never the point.
CFOs at $10M to $100M consumer brands lose time every week to model maintenance: importing, rolling forward, fixing broken references, answering the third Slack ping asking what DTC gross sales were last month. That tax goes toward zero.
Three things compound from there:
- Faster close. Imports and roll-forward that used to take a morning take a conversation.
- Multiplayer mode. Anyone on the team can ask the model a question. The analyst stops being a help desk.
- Always-current data. Models are always one prompt away from fresh.
The hours you reclaim aren't for more dashboards. They're for the modeling, the scenario work, and the decisions that move the business.
Why this works the moment you click ‘chat’
There are plenty of ways to put AI in a spreadsheet. A general-purpose chatbot. A formula helper. A planning platform with an "AI" badge stuck on the side.
The reason the SmartModel™ Agent works from the first prompt is that the heavy lifting was done before you opened it.
The Drivepoint platform already maps your data sources, normalizes them, and connects them to your model. We've spent years working with ~100 consumer brands, watching how they actually structure their forecasts, where the patterns repeat across the portfolio, and what good looks like at $5M, at $50M, at $500M. That's the context the agent runs on.
Frontier LLMs do what they're already good at: writing formulas, fixing broken references, navigating sheets, explaining the math. The Drivepoint harness does what they aren't good at on their own: live data, persistent context, business-specific guardrails, the structural knowledge of how a CPG model is actually wired, and the team dynamics that turn a one-person hack into shared infrastructure.
That's the combination. That's the last 20%, done.
Where this goes next
The SmartModel™ Agent is the foundation, not the finish line.
What's coming is a widening surface area: more pre-loaded skills, deeper integrations into the systems where the rest of the work lives, tighter loops between the model and the operational decisions it informs. The team has a long list of "can the agent do X?" prompts to answer. We'll keep answering them by shipping.
The bigger arc is the one we've been clear about from day one. The model is the heart of how a consumer brand makes decisions. Putting an agent inside the model isn't a feature. It's the foundation for what finance teams look like next.
Your model stays your model
Most planning platforms are asking finance teams to migrate. Move your model into our platform. Adopt our planning environment. Retrain the team on our way of doing things.
We don't believe that's how this gets won.
Your model stays your model. The agent just makes it faster, smarter, and answerable to anyone on the team. Single-player turns into multiplayer. The analyst stops being the help desk. The CFO stops losing Mondays to maintenance.
That's Drivepoint SmartModel™ Agent. We're glad it's here. We're more glad about what gets built on top of it.
