Technology decisions are often framed as product comparisons: cloud or on-premises, subscription or perpetual licence, commercial or open source. Those labels matter, but they do not decide whether a system will work for a particular organisation.
A useful decision starts with the work, constraints and responsibilities—not the product demonstration.
Define the outcome before the solution
Describe what should improve in operational terms. For example:
- staff can retrieve a customer record without searching several systems;
- a new starter receives the correct access on their first day;
- important files can be restored within an agreed period; or
- management can see the status of work without requesting a manual report.
This makes it easier to reject features that look impressive but do not address the problem.
Understand the current process
Document who performs the work, which information is used, where delays occur and which exceptions consume the most effort. A new system applied to an unclear process can make the confusion faster rather than remove it.
Ask whether configuration, training, ownership or a smaller integration would solve the problem before replacing an entire platform.
An AI and automation readiness assessment can help identify useful opportunities and check the data, security and integration requirements before you commit to implementation.
Compare total cost, not the headline price
Include:
- licences, storage and usage charges;
- implementation and data migration;
- devices or connectivity;
- integration and customisation;
- training and temporary loss of productivity;
- support, security and administration;
- future upgrades; and
- exit, export and replacement costs.
Free software still requires operation and support. A subscription can reduce capital cost while increasing long-term dependency. Neither model is inherently right or wrong.
Identify ownership clearly
Every service needs somebody responsible for accounts, configuration, supplier relationships, updates, backup and user support. If those responsibilities sit with a provider, the agreement should say so. If they remain internal, ensure the organisation has the time and access to fulfil them.
Avoid systems controlled through an employee’s or supplier’s personal account. The organisation should own its domain, data, tenant and administrative access.
Test assumptions on a manageable scale
A pilot should answer specific questions: does the integration work, can users complete the task, is reporting adequate, and can support resolve predictable failures?
Define success and failure before the pilot. Stopping or changing direction after evidence contradicts an assumption is not wasted work; it is less expensive than forcing a poor choice into production.
Plan security and continuity from the beginning
Ask how identities are managed, whether MFA is supported, how access is removed, where data is processed, and what logs are available. Establish backup and restoration responsibilities rather than assuming the platform or supplier covers them.
Consider what happens when the internet, provider or integration is unavailable. The required fallback depends on the importance of the process and the acceptable interruption.
Understand how to leave
Before committing, confirm:
- how information can be exported and in which format;
- whether configuration and audit records are included;
- how long export and deletion take;
- which fees apply at termination;
- whether another provider can take over; and
- what stops working when the contract ends.
An exit plan is not pessimism. It is evidence that the organisation, rather than the current supplier, remains in control of the decision.
Review the result after implementation
Return to the original outcomes. Measure whether time, errors, risk or visibility actually improved. Capture unexpected support effort and remove old systems that are no longer needed.
Good technology decisions are rarely about predicting the future perfectly. They are about making assumptions visible, testing the important ones and retaining room to change course when the evidence changes.