For the last 30 years, ERPs have been the system of record and, for many core processes, the system of control for finance and operations. It standardised processes, enforced approvals, and helped organisations run at scale – but it also hard‑coded the organisation itself. Workflows became screens, screens became roles, and roles became operating models. The way people worked bent around the way ERP was configured.

Today, many ERP modernisation programmes are still framed as technology upgrades: move to cloud, remove customisations, and improve reporting. However, that lens is rapidly becoming out of date.

AI is pivoting the conversation because it does not just optimise transactions; it changes how work gets done and where control sits day to day. More work will likely be initiated and guided through conversational interfaces, with ERP screens increasingly used for exception handling and specialist tasks. AI will translate these conversations into compliant execution across ERP, CRM, HCM, procurement, and data platforms. While ERP becomes less visible at the front end, it becomes more important as the stable system of record and control that agents depend on.

The next wave of enterprise technology strategy will be less about “which ERP” and more about where intelligence lives, how decisions are orchestrated, and how trust is maintained at scale.

In practice, enterprises are converging on three emerging patterns for where intelligence and control live.

1. ERP + Agents

• What is it: ERP stays the system of record and the system of control. Users will decrease navigating ERP screens and instead interact with an agentic interface that translates intent into compliant ERP transactions and cross-system actions, with full audit trails.
• Where it is good: Established ERP estates, regulated environments, organisations that cannot compromise controls and auditability.
• What must be available: Strong APIs and integration hooks, clean roles and entitlements so agents have clear identities, disciplined master data management, rich workflow telemetry and logging, and robust exception handling.
• Example today: A sales rep engages in Salesforce’s AI Agent, Agentforce, requesting it to renew a customer and check credit. The agent updates CRM records and triggers downstream actions in ERP via APIs, such as creating or updating a sales order, applying credit holds, or enforcing delivery blocks within the ERP. The ERP still retains the core controls and audit trails, even as the experience layer shifts to a conversational interface.

2. Policy + Orchestration

• What is it: A central orchestration layer becomes the “traffic controller” for work. Business events trigger workflows; policy‑as‑code defines what is allowed; agents execute steps across ERP, CRM, HCM, procurement, and specialised apps, rather than embedding process truth inside one system.
• Where it is good: Complex application landscapes, frequent change, M&A-heavy environments, teams that need agility without constant ERP reconfiguration.
• What must be available: A clear shared business event model, robust orchestration/workflow layer, policy-as-code governance, reliable integrations, clear product ownership for end-to-end journeys and controls.
• Example today: A buyer initiates a procurement request which is captured in ServiceNow as a workflow case. ServiceNow’s AI Agent orchestrates a policy-checked process across Finance for budget validation, Legal for contract review, and Procurement for vendor checks. Once the required controls are met, the AI agent can execute and track the end-to-end actions across systems, such as creating the PO in the ERP, initiating the contract workflow in the electronic signature platform, and updating the supplier record in the procurement platform, while providing a single view of status and exceptions.

3. Data Products + Supervised Agents

• What it is: A governed data product layer becomes the shared enterprise context and the place where decisioning happens. Agents operate on trusted data products and a semantic layer to detect exceptions, recommend actions, and trigger execution in operational systems, while a human control tower may supervise quality, risk, and accountability.
• Where it is good: Enterprises scaling AI across domains where cross-functional optimisation is the prize (working capital, service levels, margin where no single transactional system has the full picture), where you want AI to improve decisions (not just automate tasks)
• What must be available: Mature data product operating model, semantic layer, LLMOps / agent ops evaluation and monitoring, human control tower, auditability and incident response for agents.
• Example today: A company's billing, ERP, and CRM data is unified in Snowflake as governed data products with semantic views defining key metrics like receivables aging and DSO. An AI agent, Snowflake Cortex, analyses this trusted context, flags a working capital anomaly such as overdue receivables, alerts the finance control tower with supporting evidence and recommended next actions. Following human review and approval, the agent triggers the appropriate collections workflow.

Most organisations will end up with a mix of these three patterns. The winners will make deliberate choices by domain, with clear boundaries for control and accountability. They will also invest early in robust risk and governance models – for example, policy‑as‑code for agents and clear decision rights – alongside disciplined change management, because scaling AI‑driven workflows is as much an operating model shift as it is a technology upgrade. Without clear guardrails, decision rights, and a structured adoption approach, even the best architecture will struggle to deliver value safely and sustainably.

If you are designing an AI‑ready enterprise technology strategy and want to understand where to anchor control – in ERP, in orchestration, or in a data‑centric decision layer – connect with our Technology Advisory practice at Tenet. We support you from decision making through to execution, helping you weigh trade-offs by domain, identify the highest value starting points, and maximise the return on your technology investments while maintaining the controls and governance your business relies on.

This article has been created based on the experiences and insights from both Tenet Advisory & Investments and Slate & Axion.