Having worked for several companies over the years, one thing continues to surprise me: how often the same small operational problems come back. Different company, different industry, different people, different IT landscape, but remarkably similar frustrations.
These are rarely the big problems that make it onto a strategic roadmap. They are usually too small for that. None of these problems sounds particularly difficult to solve. And that is exactly why they often remain unsolved. Individually, they rarely justify a major software project or significant budget. Collectively, however, they consume a surprising amount of time.
A very typical example is time tracking. Many companies need employees to register time, but requirements often differ between departments, projects and types of employees. Another is expense management, where employees still regularly struggle with paying something on behalf of the company and getting reimbursed afterwards.
The same applies to subscription management. Companies increasingly have dozens of recurring software and service subscriptions spread across departments and employees. Knowing what subscriptions exist, who owns them, how much they cost, when they renew and whether they are still being used is surprisingly difficult. I wrote about this growing subscription jungle already several years ago in "Manage the New Jungle of Your Subscriptions" (https://bankloch.blogspot.com/2020/08/manage-new-jungle-of-your-subscriptions.html?utm_source=chatgpt.com).
Even something as basic as planning a meeting remains unnecessarily difficult, particularly when external people are involved. Internally, calendar systems can show availability, but as soon as people from several organizations participate, we are back to exchanging possible dates, using scheduling polls or sending multiple emails. The calendar should theoretically be one of our most integrated productivity tools, but there are still important gaps, something I discussed in "Your Digital Agenda: Cornerstone of Your Digital Life" (https://bankloch.blogspot.com/2020/04/your-digital-agenda-cornerstone-of-your.html?utm_source=chatgpt.com).
Other examples are even smaller. Organizing a group gift for a colleague, for example, usually results in somebody collecting money, maintaining a list of who has paid and sending reminders. It is a tiny process, but one that is repeated in almost every company. I wrote about this particular example in "Group Gift: A Precious Tradition Being Digitized" (https://bankloch.blogspot.com/2020/12/group-gift-precious-tradition-being.html?utm_source=chatgpt.com).
Similar gaps exist around employee availability and employee data management, where information about holidays, working locations, contact details, skills, responsibilities or other employee information is often distributed across HR systems, calendars, spreadsheets and collaboration tools. Read also my blog "A good employee directory - Why not part of the standard productivity tool set?" (https://bankloch.blogspot.com/2020/12/a-good-employee-directory-why-not-part.html") for more info.
And then there is task management, probably one of the clearest examples of functionality that exists everywhere and yet somehow remains fragmented. Outlook, for example, already allows emails to be converted into tasks and followed up, but these capabilities are still relatively rarely used in a structured way. Getting a simple overview of all tasks of a team can still be surprisingly difficult. Companies introduce Jira, Planner, Trello, Asana or other tools, while many tasks continue to live in emails, meeting notes, spreadsheets and people’s heads. See my blog "Streamlining Task Management: Can One Tool Fit All?" (https://bankloch.blogspot.com/2024/11/streamlining-task-management-can-one.html) for more info.
For almost every one of these problems, a software solution exists somewhere. But introducing a dedicated solution for every small operational gap quickly becomes unrealistic. Each new SaaS application comes with a subscription, implementation effort, user management, security reviews, training and support. More importantly, it adds another application to an already fragmented landscape. Before long, employees are switching between dozens of tools, often entering the same information multiple times because those tools are not properly integrated.
So companies tend to take the pragmatic route. Someone creates an Excel file. Another person builds a SharePoint list. A team starts maintaining something in Teams or Google Sheets. Perhaps somebody discovers an overlooked feature in Outlook or another application that is already available. Initially this works perfectly well. The problem is that a temporary workaround has a tendency to become a permanent business process.
Excel is probably the best example. It is incredibly flexible and almost everyone knows how to use it. For a small problem, creating a spreadsheet is often much easier than requesting a budget, selecting software and going through IT. But then the spreadsheet grows. More people start using it. Copies appear. Macros are added. Someone needs information from another system, so an export is added. After a few years, nobody is entirely sure which version contains the correct dataor how the spreadsheet actually works. What started as a clever solution to avoid unnecessary complexity has itself become complex.
There is another reason why these gaps are difficult to solve. They are often very company-specific. A standard SaaS product might solve 80 percent of the problem but miss the 20 percent that actually matters for a particular process. Sometimes the functionality is already available through configuration or customization, but nobody knows it. Because the application is relatively small, there is no real owner and certainly no internal expert who understands everything the tool can do. Buying another tool then seems easier than improving the one already there.
Generative AI is now changing this equation again. With AI-assisted development, or "vibe coding" as it is called, creating a small custom application has suddenly become much easier. An employee who would previously have built an Excel sheet can now create a simple web application. With surprisingly little effort, it can have a proper interface, workflows, validations and even integrations. For many of these small software gaps, the result can be significantly better than yet another spreadsheet.
This is a very interesting evolution, but it also introduces a new risk. The fact that we can build software much more easily does not necessarily mean that we should build more software.
When development was expensive, there was a natural pressure towards reuse. If one department invested substantial time and money in developing an application, other departments were encouraged to use the same solution. Today that barrier is disappearing. If every department can create its own small application in a few days, why bother aligning with another department? Why compromise on requirements when you can simply build exactly what you need?
The consequence could be a new generation of shadow IT. Instead of hundreds of uncontrolled Excel files, companies may end up with hundreds of uncontrolled small applications. And unlike a spreadsheet, an application introduces additional questions. Who maintains it when something breaks? Who updates the libraries it depends on? Who checks its security? Where is the data stored? Who manages access rights? What happens when the employee who created it leaves the company? And how does it exchange information with the rest of the organization?
The initial development cost may have fallen dramatically, but the lifecycle costhas not disappeared. In some cases it has simply become less visible.
This is why I believe the real opportunity created by AI is not necessarily to let everyone build more standalone applications. It is to make existing integrated platforms much easier and cheaper to extend.
Most companies already have several platforms that could potentially cover many of these small gaps: ERP systems such as SAP, Oracle or Odoo, CRM platforms such as Microsoft Dynamics, HubSpot or Salesforce, productivity suites from Microsoft or Google, HR platforms, and increasingly even banking applications. These platforms already have users, authentication, permissions, data models and integrations. Extending them therefore starts from a very different position than building another standalone tool.
ERP systems are particularly interesting in this context because many of the small operational gaps eventually come back to shared company data. Employees, customers, suppliers, projects, expenses, working hours, invoices, products and subscriptions are not isolated concepts. They are connected. The more applications that independently maintain copies of this information, the more reconciliation becomes necessary.
And reconciliation has a cost that companies often underestimate. An employee spending five minutes copying information from one system to another does not generate an IT invoice, so the cost remains largely invisible. The same is true for colleagues comparing two spreadsheets, correcting inconsistent data, searching for the latest version of a file or asking somebody on Teams whether information is still up to date. Multiply those small actions by hundreds of employees and hundreds of working days, and suddenly the "cheap" solution is not so cheap anymore.
Platforms such as Odoo are interesting examples of a possible middle ground. They combine many business functions within one environment while remaining relatively easy to extend. The open-source model also makes deeper customization possible when required. Combined with AI-assisted development, this could make relatively small extensions economically viable without automatically creating another isolated application and another copy of company data.
This does not mean that every process should be forced into an ERP system. Sometimes a specialized SaaS product really is the better solution, and sometimes an Excel file genuinely is all that is needed. The objective should not be centralization for the sake of centralization. The question should rather be whether a small local solution is still local once it starts depending on shared data, multiple employees or recurring business processes.
AI will make software creation increasingly accessible. That is undoubtedly a good thing. Many frustrating gaps that were previously too small to justify development can finally be addressed. But companies will need to resist the temptation to treat the lower cost of creating software as proof that the total cost of software has disappeared.
Perhaps the next step in digitalization is therefore not an explosion of thousands of individually generated applications. It could be the opposite: using AI to finally close all those small gaps inside a coherent landscape, where applications share data, common components and governance.
For years we accepted spreadsheets, manual re-entry and endless small alignments because solving each problem properly seemed too expensive. AI is changing that calculation. The opportunity now is to solve those problems without replacing spreadsheet sprawl with application sprawl, and to use the technology to make our existing platforms fit the way companies actually work.
Comments
Post a Comment