Manufacturing is past the pilot question. Three surveys published this month agree: adoption is scaling, value is real, and the bottleneck has moved from models to data and people. Here is what the numbers say, and what it means for our team.
The numbers this week
1. 84% of manufacturers now report measurable value from AI in operations, with average improvement potential of about 20% across core KPIs. Adoption is strongest in quality (62%), production (57%), and logistics and supply chain (49%). — Deloitte Germany, AI in Manufacturing Survey 2026
2. More than 70% of manufacturers now use generative AI tools like ChatGPT or Copilot, up 24 points since 2024. Two in three use or plan to use agentic AI. The top blocker: poor data quality, named by 62.3% of respondents. — Manufacturing Leadership Council Industrial AI Survey
3. 44% of organizations report AI scaling across the enterprise, up from 38% a year ago. At large companies, agentic AI in production jumped from 27% to 40% in one year; smaller firms stayed flat at 22%. — McKinsey State of AI 2026
4. 77% of finance organizations now employ AI, and forecasting adoption rose from 58% to 76% year over year. But only 35% say they are effective at measuring AI ROI, and just 14% work from a defined AI strategy. — Protiviti Global Finance Trends Survey 2026
Why it matters for Justrite
The gap between large and small organizations is widening, and it is not about model access. Everyone has ChatGPT-class tools. The difference is data readiness and process discipline. The companies scaling agents have clean, governed data. The ones stuck in pilots are fighting spreadsheets, silos, and the 62% data problem. We get to choose which side of that line we land on — and the choice is mostly about data hygiene, not technology.
Tool watch: SAP pushes agents into finance
SAP's "Autonomous Enterprise" rollout adds a financial closing assistant that surfaces bottlenecks, automates postings and reconciliations, and resolves discrepancies in real time — plus assistants for planning, billing, AR, governance, tax, and treasury. It follows similar moves from Oracle and Workday. (CFO Dive)
My take: ERP vendors are making agentic AI a standard feature, not an add-on. The question shifts from "should we use agents?" to "which of our processes are ready for them?" The honest answer for most teams: the ones with clean data and clear rules.
What they won't tell you
Vendors demo agents on clean data. They don't tell you that 62.3% of manufacturers name poor data quality as their top blocker — before any model is involved. They don't tell you that most manufacturing AI spend sits in analytics and reporting, the lowest-ROI spot, while transactional workflows like invoicing, reconciliation, and three-way matching stay manual. (Yooz 2026 AI in Finance Report) And they don't tell you that CFOs' most pressing near-term concern is not technology or budget — it is building AI talent inside the finance team. (BCG AI Radar 2026) The model is never the bottleneck. Data and people are.
This week's move
Pick the longest step in your month-end close. Write down what it involves. Open ChatGPT Enterprise. Upload the raw data for that step. Ask it for the first draft — reconciliation notes, variance explanation, whatever it is. Keep a human in the loop. Time it. That is your pilot.
The full prompt library lives on the AI@Justrite SharePoint site, Prompts folder — copy, adapt, and run.
From our shop floor
Last week in an operations review, someone asked why we can't just "switch on" AI for planning. The honest answer: the models are ready. Our data is not — yet. That is the work.
Discussion question
Which report does your team rebuild from scratch every month? Reply to this email. If three people name the same report, we will automate it next.
— Remus Samoila, IT Analyst, Automation & AI