Stabilising and recovering performance after an ERP go-live

What distinguishes leading recoveries is the introduction of AI‑enabled capabilities at critical points in the process. Combining structured methods with data‑driven insights shortens time to clarity and reduces the operational burden on teams already stretched by go‑live.

Phase 1: Mobilisation & immediate stabilisation

A key component in phase 1 is to enable decision-making and governance. The reason behind this is to consider the people and the organisational stress inherent at the time of project initiation. When we are faced with post-go-live optimisation or recovery projects, it is typically at the end of a very stressful and busy period for the whole organisation in both IT and business. As such, catering and enabling a structured outside-in view on priorities and decisions to be made is very beneficial. However, it must always start with a people-centric point of view.

This phase establishes the control tower for recovery and creates the operating rhythm that defines the next 6–12 weeks. We typically start by establishing:

A clean slate. Meaning that we conduct an unbiased, but experienced, review of key priorities to solve short-term. Free of judgement from past and current errors and assumptions. Additionally, we focus on governance through clear roles, issue owners, escalation paths, daily standups, weekly sprints with multiple face-to-face interactions to ensure correct prioritisation and progress.

Furthermore, transparency and actual status on day-to-day or hourly basis is hard. We focus on prioritising operational indicators such as order fulfilment, inventory accuracy, production throughput, and financial postings.

To enable fast decision-making, we ensure that a cross‑functional task force is empowered through representatives from operations, supply chain, finance, IT, data, and functional leads, with real business insights and the mandate to make decisions in real-time.

We rarely utilise AI at this step, as the key task in mobilisation is to re-initiate the organisation – i.e., its people – and determine the right sequence of initiatives. However, if existing tools can provide insights into bottlenecks, process constraints, or operational KPIs, they may be considered. The value is not in the tool itself, but in removing the weeks spent manually tracing or observing where transactions are actually failing. It cannot replace physically spending time on the shop floor, in the planning department, or in the warehouse.

Phase 2: End‑to‑end assessment and prioritisation

When the immediate operational bleeding is slowed (ideally stopped entirely), the next step is a structured diagnostic across the entire value chain with selected deep dives into the identified key areas.

At Implement, our approach includes:

• End-to-end walkthroughs of Order-to-Cash, Plan-to-Produce, Procure-to-Pay, and Record-to-Report. Depending on the preferred process framework, we have our own industry-based process landscapes and operational models.
• Process mapping against actual system behaviour to identify and highlight gaps, constraints, bottlenecks within people, processes, and/or the system.
• Master data validation and walk-through audits of critical elements identified e.g., BoMs, routings, pricing, business partner records.
• Integration assessments across WMS, MES, CRM, BI, and finance, if relevant.
• Prioritisation via impact/effort scoring to determine first‑wave improvements.

In Phase 2, we often see a better use-case for utilising the capabilities of AI, including:

• Process mining for bottleneck detection to automatically reveal slow steps, rework loops, and variant explosions, while simultaneously supporting both clinical diagnosis and prioritised improvement. This can be done, e.g., by using SAP Signavio, but Best-of-Breed programmes can also be utilised, depending on preferences or fit.
• AI‑driven master data anomaly detection, which flags outliers in critical master data such as BoM structures, unit measurements, pricing conditions, material statuses, and business partner records.

When we look at the most advanced organisations in AI utilisation, AI is used to leverage process intelligence to quantify deviations from best-run processes and benchmark performance against industry standards before committing to any changes. With the new capabilities of individual AI tools, it is also possible to build process simulations and volume validations prior to implementation, allowing organisations to assess the impact of potential changes.

Phase 3: Stabilisation & first‑wave improvements

At this point, organisations can begin implementing improvements that remove friction systematically and are more long-term in nature.

Implement’s first‑wave stabilisation focuses on:

• Redesigning planning models to align with actual lead times, production constraints, and demand signals.
• Simplifying or clarifying warehouse and production execution flows to reduce errors and manual interventions.
• Improving integration reliability and automating error handling.
• Enhancing user guidance through targeted trainings, playbooks, and communication.
• Extending governance routines beyond firefighting, for example through Global Process Owners and the organisational adaptation of governing more than one plant or site via template-based ways of working and clear ownership.

Further and long-term stabilisation relies heavily on continuous monitoring of cycle times, support-tickets, process deviations, and task throughput. Identifying where workflows take longer than intended is a core principle of performance optimisation frameworks, especially as we dive further into the complex processes and interdependent constraints across business functions.

AI use in Phase 3

When moving further away from firefighting and short-term stabilisation, we can utilise AI to further expand and analyse the reasons behind the initial decrease in performance. That could be achieved by using root‑cause analysis through machine learning to identify systemic drivers behind recurring failures, e.g., correlation between demand variance and MRP exceptions. Or to execute AI‑supported planning simulations by using historical data to model alternative planning parameters (lot sizes, reorder points, safety stocks etc.).

These tools support a shift from reactive fixes to proactive stabilisation – exactly what is key in Phase 3 and a critical component in moving from short-term fixes and stabilisation to long-term optimisation.

In one specific client case, we managed to take a customer from a forecast accuracy of 50% to 80%, while increasing their delivery reliability from 70% to 98% and reducing their net working capital by 10%. All within the Phase 3 initiatives. After which, the team was reduced and the focus shifted to more long-term template and roll-out work to avoid similar situations in the future.

Phase 4: Long‑term optimisation & capability building

Once the organisation exits crisis mode, the focus shifts to embedding new ways of working and creating a sustainable improvement engine. Our model for long‑term optimisation can vary greatly depending on the industry, the customer, and their strategic goals, both in business and within IT. However, typical steps include embedding governance routines such as KPI reviews, cross‑functional decision boards, and continuous improvement forums into the organisation and daily business.

Strengthening process ownership and documentation, both by the establishment of a process organisation linked to the decision-making process and technically to document the agreed processes. The latter can be accelerated using process documentation tools and automated workflow tools.

AI use in Phase 4

The use of AI tools in Phase 4 depends greatly on the company and industry. And as this targets more long-term improvements and optimisations, it likewise depends greatly on the individual company’s goals and aspirations within the area. However, when aligned with a pre-existing strategy, this can involve AI agents monitoring patterns in process flows and supporting exception handling, alongside more standardised AI capabilities, for example within SAP. Predictive functionality can be found in solutions such as SAP IBP, AATP and EWM, helping to prevent stock-outs, improve stock levels, and provide timely business alerts.

Additionally, the further we progress through the phases, the closer we get to strategic business decisions and impact. And that means moving further into the territory of more strategic and tactical decision-making on behalf of the overall business and away from recovery and post-go-live optimisation, which is a natural consequence of this type of project. We just need to recognise the grey zones and overlap, so that we may ensure a clear understanding of the business implications and consequences.

The main takeaway

ERP stabilisation is both a challenge and an opportunity. We can, however, identify several distinct markers to the organisations that thrive after go‑live. Those organisations are characterised by the ability to accept instability as normal, at least as a temporary state. They apply structure before applying speed. Often with assistance from externals, recognising that they are biased and impaired by past events and attempts to resolve the issues. They use facts, data, and, increasingly, AI to drive decisions and focus on the work. They spend time on cross‑functional alignment rather than siloed fixes. And finally, they prioritise action, especially in Phases 1 and 2 over documentation, governance, and capability building. Not that the latter are not important – they are – just not as important in the short-term recovery context.

The main takeaway should be that if you and your organisation find yourselves struggling after an ERP implementation, you are not alone and certainly not the first. The good news is that there is also a proven way out of this temporary setback – and even an upside. For organisations that do struggle, we see faster recoveries and a return to normal once the problems are recognised and addressed. If you have read this far, it is already a testament to that and a first step towards returning to normal. So, if you need help or simply wish to spar on how to avoid the pitfalls of ERP implementations, please reach out.