Accelerators
›Purpose-built LLM systems, multi-agent reasoning workflows, and domain-adapted models.
Universal semantic metric layers, natural language querying, and executive intelligence.
Mathematical constraint solvers (MILP, CP-SAT) for scheduling, allocation, and routing.
Pragmatic ERP modernisation, event-driven API mesh, and custom operational applications.
Industry Trends in AI & Analytics
Practical benchmarks on what actually works when moving AI from pilot to production.
[VISUAL V1: hero illustration - PLACEHOLDER, image not yet supplied]
Your sales team drafts emails with ChatGPT. Meanwhile, orders still arrive as WhatsApp voice notes, and the stock count lives in a spreadsheet someone updates on Fridays. That gap is common. A CII-ISB study of 2,919 MSMEs, as reported by Business Standard, found 65.8% use AI tools but only 21% use ERP or customised business software. That is the wider MSME pattern. Larger firms often show a softer version of it: an ERP exists, but key decisions happen outside it. Our view is simple. AI returns money when it can read and act on reliable operational data. Without that, it stays a writing assistant.
[VISUAL V2: bar pair, 65.8% vs 21% - PLACEHOLDER, image not yet supplied]
ERP readiness for AI: why tools alone rarely show up in the P&L
Chat tools help people write faster, and that is useful. But margin moves when decisions on stock, pricing, purchasing and production change. Those decisions need current, trusted numbers from the system that runs operations. If that system is a mix of spreadsheets and message threads, AI has nothing dependable to act on.
The same study, as reported by Business Standard, shows where AI use sits today: business communication (34.9%), content creation (22.3%) and training (18.9%). These are personal productivity gains, not operating ones. We have written separately about why AI pilots do not scale; missing operational data is one of the main reasons.
What a data backbone means in plain terms
A data backbone is one system of record for items, customers, suppliers, stock and orders. Its master data is clean, transactions are captured when they happen, and other tools connect to it so everyone reads the same numbers. For most mid-market firms, that system of record is the ERP, used properly.
[VISUAL V3: layer diagram - PLACEHOLDER, image not yet supplied]
System of record versus systems of conversation
WhatsApp, email and Excel are where people talk about the business. The ERP is where the business is recorded. Trouble starts when an order is agreed in a chat, adjusted in a spreadsheet and entered into the ERP days later, if at all. Each copy drifts, and no one is sure which number is right.
Why "we have an ERP" is not the same as "we are ready"
Many firms own an ERP but use it mainly for invoicing and statutory reports. Stock is updated weekly, purchase approvals happen over the phone, and the real planning sheet sits on one person's laptop. The licence is paid for, but the data AI would need still lives elsewhere.
Is your ERP ready for AI? A 7-point check
Your ERP is ready for AI when it holds current, trusted data and the decisions you want to improve already run through it. This check is our own rule of thumb from project work, not survey data.
[VISUAL V4: checklist card - PLACEHOLDER, image not yet supplied]
Stock and orders are recorded in the ERP the same day.
Each item has one code and one unit of measure.
You can name the owner of item, customer and supplier data.
Reports come from the ERP, not from exported spreadsheets.
Other systems connect through APIs or scheduled feeds, not copy-paste.
Approvals for purchase, credit and price happen inside a workflow with an audit trail.
You can say which decision AI should improve first, and how you will measure it.
If you answered yes to fewer than four, your first project is data and process, not AI. That is good news: it is usually smaller and faster to pay back than people expect.
Example: a food-service distributor on Microsoft Dynamics NAV
A food-service distributor in India did not replace its ERP. We connected its live stock data to dashboards and forecasts, then let forecasts raise reorders inside Microsoft Dynamics NAV. The distributor had been losing margin to perishable shrink and to next-day reports that forced emergency buys.
What we built
We built live stock dashboards with a freshness view for every perishable item, added demand forecasts, and set up forecast-triggered reorders written back into Dynamics NAV.
[VISUAL V5: flow from live stock to purchase order in Dynamics NAV - PLACEHOLDER, image not yet supplied]
What changed
The distributor saw 18% less perishable wastage, a 96% fill rate and Rs 1.2 crore in added margin.
The dashboards are the kind we build in business intelligence and analytics work. For why the forecast alone was not enough, see an accurate forecast can still drive the wrong decision.
Modernise, extend or replace? Choosing the first step
Most mid-market firms should extend the ERP they have before replacing it. Clean the master data, capture transactions in the system and connect the tools around it. Replace only when the system is unsupported, cannot integrate, or forces core work outside it.
Your situation | Likely first step |
|---|---|
Duplicate item codes, unclear data owners | Clean master data |
Data is good but trapped in one system | Connect and integrate |
Orders and approvals run on WhatsApp and spreadsheets | Add workflow apps |
ERP is unsupported or cannot connect | Upgrade or migrate |
Our ERP implementation and modernisation work covers each of these, on SAP, Infor LN and Microsoft Dynamics. Where the gap is a manual step, enterprise applications can replace the WhatsApp group or the spreadsheet with a proper workflow.
Where AI fits once the backbone is in place
On top of clean ERP data, AI can forecast demand, flag exceptions, suggest reorders and answer questions about stock in plain language, with people approving the actions. The AI is the same kind of tool your team already uses. The difference is that it now works from numbers you trust.
Two places to start: let managers ask questions of your data without SQL, or apply predictive analytics with machine learning to one decision with clear money attached.
FAQ
What does ERP readiness for AI mean? Your ERP holds current, clean data on items, stock, orders, customers and suppliers, records transactions when they happen, and connects to other tools. AI can then read and act on the same numbers your team trusts, instead of guessing from exported spreadsheets.
Do we need a new ERP before using AI? Usually not. Many firms get further by cleaning master data, capturing transactions in the existing ERP and adding integrations. Replacement makes sense when the system is unsupported, cannot connect to other tools, or forces core work outside it.
Why are AI tools not improving our margins? Most AI use today is writing and communication. Margin moves when stock, purchasing, pricing and production decisions improve, and those depend on reliable operational data. Without it, AI has nothing dependable to act on.
How many Indian MSMEs use ERP compared with AI? A CII-ISB study of 2,919 MSMEs, as reported by Business Standard, found 65.8% use AI tools while only 21% use ERP or customised business software. It describes MSMEs broadly, not any single segment.
Can AI work with Microsoft Dynamics NAV? Yes. In one of our projects for a food-service distributor, forecasts triggered reorders written back into Dynamics NAV, alongside live stock dashboards. The result was 18% less perishable wastage, a 96% fill rate and Rs 1.2 crore added margin.
What is a sensible first project? Pick one decision with clear money attached, such as reordering perishables or chasing late orders. Make sure its data is clean in the ERP, show it on a live dashboard, then add a forecast or alert that a person approves.
Not sure where your ERP stands? Talk to us about your stock, order and master data. Book a discovery call
Get a 7-point ERP readiness check
KEEP READING
Explore more perspectives on AI, analytics, and enterprise intelligence.
