It's tempting to think of the process of implementing an AI-powered ERP system as a journey, with a set of highly attractive rewards waiting like a pot of gold at the end of the implementation rainbow.
The potential rewards are real, of course. Deployed correctly, AI can help you:
-
-
make more accurate business forecasts
-
solve supply chain issues
-
prepare for disruptions and disasters
-
boost productivity
-
refine your pricing strategy to increase your margin
-
keep your customers happy
-
reduce equipment downtime
-
… and much more.
And in many respects, the journey metaphor is perfectly valid. Like a journey, it can be broken down into distinct stages: the preparations you need to make before setting out; the time you need to spend mapping out your path and getting good directions from a trusted guide; the execution itself and the pitfalls you need to avoid along the way; and, of course, the periodic assessments you need to do to gauge your progress and make sure you're still on the right path as you move forward.
But it's important not to fall into the trap of thinking that successfully implementing your ERP is the end of the journey. In fact, it’s the beginning of a new journey — one that requires ongoing diligence, collaboration, and company-wide participation to make sure that you're able to truly reap the rewards of your investment in AI.
Setting Up for Success: Don't Assume the Data
Before we talk about the post-implementation journey, it can't be stressed enough that your eventual success depends on how well you plan and prepare — particularly in terms of your data.
Whether you're using sophisticated analytics or just doing two-dimensional reporting; whether you're doing machine learning or RPA, data is the lifeblood, and you can't afford to take it for granted.
In other words: You can't just assume the data.
In fact, assuming the data is one of the earliest potential pitfalls. Heading into a project, one of the biggest blind spots is often the assumption that the data is already shipshape and ready to go. Clients often assume that:
-
-
The data is going to be available.
-
The data is going to be error-free.
-
The data is going to be consistently formatted.
-
The data is going to be compatible.
-
The data is going to be complete.
-
All of the data is going to be what they need.
-
In reality, there's a lot of work that needs to go into defining what data is needed, what needs to be done to clean it up and get into the right format, and how it's going to be managed in an ongoing way. If you avoid assuming the data, you'll increase your odds of eventual success.
Curating Your Data Set
Your goal is to arrive at what's known as a curated data set: one that's valid and ready to go. That's not to say that the data and your approach might not change further down the road — but for the POC, you've got a set of data that's been vetted and standardized and harmonized.
That often requires a fair amount of engineering, and building what can be thought of as Rosetta Stones to convert and consolidate the data, so that you come up with a common data set that's clean and qualified.
As an example of what can go wrong if you don't have a curated data set: Columbus did a machine learning experiment for a manufacturing company that had three years of historical data, and within it there was about three weeks of data that didn't make any sense. After investigating, Columbus found that during that three-week period, a lot of the company's systems were intermittently shutting down and coming back up, resulting in data values that were way out of alignment. Those values needed to be removed from the data set in order to make it coherent and usable. That act of curation was necessary to ensure the integrity and functionality of the data as a whole.
Speaking of Data ...
It's critically important to make sure that everyone is speaking the same language when it comes to data management. It's a little like being from the Midwest and visiting Boston or New York (or the other way around). You know they're speaking English, but you might not be completely sure you know what they're saying.
In the data space, defining terms at the beginning of the project is crucial, because people often use the same terms differently without realizing it. What exactly is meant by optimization, or predictive analytics, or updating data (as opposed to, say, inserting data)? If everyone isn't speaking a common language, it's harder to get everyone rowing in the same direction — and it can lead to mistakes and misunderstandings downriver.
Five Keys to Post-Implementation Success
With that in mind, here are five ideas that can help guide you through the post-implementation landscape.
1. Moving at the right speed
Implementation doesn't have to happen overnight. In fact, sometimes it works better if you take it gradually.
So instead of installing IoT devices throughout the entire factory, for example, you might start with a couple of key pieces of machinery to see how it works. That gives stakeholders who are unsure or skeptical an opportunity to see the benefits and positive results, which in turn helps create the enthusiasm for further progress. With this approach, post-implementation for one step of the process is where you begin implementation of the next step.
2. Managing change
Having an OCM (organizational change management) plan in place is a crucial piece of the puzzle. Not having a well thought-out approach to change management is one of the biggest reasons why implementations sometimes fail.
Effective change management includes establishing the reasons for the change, so everyone understands the goals as well as the way forward, and then engaging your team in regular, ongoing communication to address issues and measure progress.
Ideally, this communication process continues into the post-implementation phase and beyond, since change is a continuous element of business in the Industry 4.0 era.
3. Ensuring adoption
AI tools don't do you any good if your team members don't embrace them and incorporate them into their daily workflows. Some team members may be inclined to stick with established routines and tools out of familiarity and habit. So it's important to communicate with team members at every level of the organization to make sure they understand how these new tools can ease their pain points, help them accomplish their own personal goals, and allow them to achieve things they might not have previously thought possible.
An essential part of this process is listening to your team members' concerns. By finding out what your employees need help with, and what anxieties or trepidations they may have about the implementation, you can do a better job of removing obstacles and smoothing the path to adoption.
4. Managing the data
In order to maintain the quality, integrity, and security of data, it's necessary to have processes in place for how data is entered, as well as checks and balances to make sure that data is used correctly and ethically. Proprietary code and sensitive data like customer information need to be respected and appropriately flagged.
Building the right security around your data from the ground floor up, including two-factor authentication at every level, is a necessary first step.
And since data is democratized in a centralized ERP system, with lots of different team members entering it, it's vital for there to be agreements in place that everyone is aware of and everyone respects, so that data is entered consistently and correctly.
5. Providing ongoing training and assessment
To make sure that your team continues to reap the benefits of AI, you need to provide regular, ongoing training so that they stay updated on all the latest changes and don’t fall behind. That means making sure that new team members get full training as well, so they aren't stuck learning in bits and pieces as they go, and that mission-critical knowledge doesn't exit the organization along with retiring or departing employees.
By the same token, your roadmap for the future needs to include regular assessments to make sure that your tech is functioning well and that your team is using it effectively. Establish key metrics, measure them, and then address issues proactively to avoid a drop-off in efficiency and productivity.
Want to learn more about how Columbus can help your organization harness the full capabilities of AI? Get in touch with us today.