Five questions assessment buyers should ask their technology suppliers

What PSI and Enovate told the Test Community Network

Published: 10/6/2026
Five questions assessment buyers should ask their technology suppliers

Ahead of the E-ATP conference, I wanted to do something more useful than publish another list of exhibitors or ask suppliers what was new on their stands.

Assessment buyers need questions that help them understand how a prospective partner thinks: how it will protect the candidate experience, respond when things go wrong, introduce AI responsibly, avoid locking customers in, and exercise professional judgement when the easiest answer is not the right one.

The Test Community Network invited 19 E-ATP suppliers to answer the same five open questions. Two organisations, PSI and Enovate, submitted complete responses. My thanks to PSI and Enovate for taking the time to provide thoughtful answers and for being prepared to respond publicly.

See the full list of sponsors and exhibitors at the E-ATP Conference 2026.

Two responses cannot represent an entire market. But these are useful examples of the kind of open, thoughtful response that buyers should expect when asking difficult questions. I hope more organisations will be willing to participate in this kind of industry dialogue in future.

The supplier responses below are reproduced in full. They have been supplied by the organisations themselves and have not been independently audited by the Test Community Network.


1. Integration and the end-to-end candidate experience

Why buyers should ask this

Assessment services are frequently assembled from several platforms. An awarding organisation may use one system for candidate management, another for test delivery, and a third for online proctoring. There is nothing inherently wrong with using specialist providers, and the assessment industry has developed some impressive API connections.

The problem is that candidates do not experience APIs. They experience interfaces, emails, installation requests, repeated questions, new usernames and unexplained hand-offs between organisations they may never have heard of. One examination may be delivered on one platform and the next somewhere else. Technically connected systems can still create a fragmented, stressful and confusing candidate journey.

What impressed me in these responses was the emphasis on creating a connected experience rather than a series of hand-offs. Both suppliers describe flexible journeys in which the underlying complexity is intended to remain invisible to the candidate. I would still like single sign-on to become the norm wherever possible.

The practical challenge for any supplier is simple: has the person selling the solution personally completed the entire candidate journey they are proposing? A supplier should experience its own hand-offs, messages, identity checks and platform changes before claiming that the journey is seamless.

The question

How do you enable organisations to create an end-to-end candidate experience across multiple systems, while minimising breaks in journey, duplicate data entry and unnecessary platform switching? Please give an example of how you have achieved this in practice.

PSI response

A candidate's journey typically moves through several connected systems: registration and scheduling, eligibility and identity checks, the testing platform and, for online proctoring, the proctoring platform, followed by scoring and reporting. We have designed that journey as one connected experience rather than a series of handoffs.

APIs and system integrations allow information such as candidate details, eligibility, booking status and delivery mode to flow between systems, reducing duplicate data entry, separate logins and unnecessary platform switching. The aim is that the complexity of the underlying technology is largely invisible to the candidate. They experience a seamless journey from registration and scheduling through to assessment and results, even when multiple systems are working behind it.

Enovate response

We try to make the assessment platform the hub of the candidate journey, while allowing the customer’s existing systems to remain the systems of record where appropriate. This means using SSO, APIs and automated workflows to avoid repeated logins, duplicate data entry and unnecessary switching between platforms.

One example is our solution for OsloMet’s national interpreter qualification schemes in Norway. It combines application, admission, scheduling, payment, candidate communication and administration in one workflow, while integrating with Norway’s ID-porten and Feide identity services and exchanging data with the National Interpreter Register.

For the Swedish national hunting examination, we took a slightly different approach. Candidates and examination leaders log in through the Swedish Hunting Register, which is already the service they know and use. From there they are automatically authenticated into Enovate’s examination platform. They therefore do not need a separate Enovate username or password, and the transition between the public register and the examination service is invisible to them.

Our approach is not necessarily to replace every surrounding system, but to make the boundaries between them as seamless as possible for the candidate and examination staff.

The buyer’s follow-up: Ask the supplier to demonstrate the whole journey, not simply the individual platforms. Also ask whether its commercial team has personally experienced that journey from beginning to end.


2. Cyber resilience and continuity

Why buyers should ask this

For assessment organisations, a serious cyber incident is no longer an abstract possibility. It is a question of when, not if, an organisation somewhere in the assessment ecosystem will face significant disruption.

The responses provide strong reassurance about continuity planning and operational support. I particularly agree with Enovate’s statement that continuity must be designed before an incident occurs. However, neither response includes a publicly shareable example of supporting a customer through a real cyber incident. I understand why organisations may be cautious about disclosure, but the sector needs greater transparency so that others can learn.

My concern is not primarily that suppliers have ignored resilience. It is that awarding organisations may not yet be having sufficiently serious conversations with them. Before an incident, customers and suppliers should sit down together and ask what happens when normal systems and integrations must be shut down.

Suppliers do not have to disclose confidential customer details. They should, however, be able to bring credible incident experience into the room, explain the lessons learned, and help customers turn those lessons into realistic continuity plans. You remember the organisations that help when things go wrong.

The question

If one of your customers experienced a serious cyber incident that disrupted their normal systems or integrations, how would you support them in maintaining critical assessment activity? Please give a publicly shareable example of how you have supported an organisation through a significant disruption or incident.

Enovate response

Our first principle is that continuity must be designed before an incident occurs. Enovate operates an ISO 27001-certified information security management system, with risk management, monitoring, backups, incident procedures and defined escalation and communication processes.

If a serious incident affected a customer or an integration, our first priority would be to isolate the affected component and maintain service continuity wherever this can be done without compromising security or assessment integrity. The specific response would depend on the nature and severity of the incident and would follow established incident and continuity procedures.

For individual customers, continuity requirements, recovery objectives and incident scenarios are agreed according to the risk profile of the assessment.

PSI response

Our delivery, proctoring, and scoring platforms operate independently of client infrastructure, so a compromise on the customer side doesn't have to mean a disruption to testing. We isolate the affected integration points and move the programme onto contingency paths so candidates can keep registering, testing, and receiving results.

In practice:

  • Scheduling: assisted booking and validated rosters when the registration system is down.
  • Authentication: our identity assurance stack (ID capture, biometric match, liveness and deepfake detection) verifies candidates without relying on the programme systems.
  • Delivery: test centre and remote proctoring act as fallbacks for each other, so sessions are rerouted rather than cancelled.
  • Results: scores are held securely and released in reconciled batches once the programme environment is verified.
  • Protection: we raise fraud monitoring thresholds and run targeted web and dark web monitoring for the duration.

In case of incident, the customer gets a named incident lead, a joint bridge at an agreed cadence, and written status updates throughout. Every contingency decision is documented, so score validity holds up afterward. Our continuity and response programmes align to ISO 22301 and ISO/IEC 27001.

The buyer’s follow-up: Ask who will sit beside you during a genuine incident, what relevant experience that person brings, and how lessons from previous incidents have changed the supplier’s continuity arrangements.


3. AI roadmaps and the risk of unpredictable costs

Why buyers should ask this

Both respondents provide strong, AI-forward answers. More importantly, both address cost predictability rather than treating AI as an unlimited source of savings.

I welcomed PSI’s honesty that AI may not reduce the overall cost of an assessment programme. Instead, investment may shift towards larger item banks, more frequent content refreshes and deeper analytics. Enovate similarly argues for defined assumptions, appropriate limits and predictable pricing rather than simply passing variable model costs to customers.

The broader point is that AI cannot be treated as another feature bolted onto an unchanged process. Replacing one manual task with an AI tool may deliver a small efficiency, but the larger opportunity comes from redesigning the entire assessment lifecycle with AI considered from the ground up. AI-native processes may change the work, the roles, the quality controls and the economics together.

No supplier can promise to absorb every future inference or processing cost. Model prices may fall while usage expands dramatically. Buyers and suppliers therefore need transparent contracts that can accommodate both increases and decreases, rather than hiding an unknown consumption bill. There may also be opportunities to move beyond activity-based charging and explore outcome-based pricing where appropriate.

The question

How do you expect AI to change your products and services over the next two to three years, and how are you helping customers benefit from AI without creating unpredictable or escalating costs as usage increases?

PSI response

How AI will change things over the next two to three years

Rather than adding AI to individual steps of an existing process, we're rebuilding the assessment lifecycle with AI considered from the ground up: content development, delivery, test security, and reporting all designed to work together.

The value isn't in any single AI-powered feature. It's in how those pieces connect across the full lifecycle, with human oversight built in at the points that need professional judgment.

On cost predictability

Adopting AI at scale doesn't necessarily reduce a programme's overall cost. What changes is where the money goes. Instead of spending on the same tasks in the same way, budget shifts towards the things AI now makes possible: a larger item bank, more frequent content refreshes, deeper analytics. We work with clients to plan that shift deliberately, so spending stays predictable as their programme grows.

Enovate response

We expect AI to become embedded in assessment workflows rather than operate as a separate tool. The greatest value, in our view, will come from task-specific AI working within structured processes for authoring, linguistic quality assurance, translation, analysis and administration.

One current example is the integration of cApStAn’s TAlly and Toledo technology into Enovate’s assessment authoring environment. AI-supported checks can flag linguistic, translatability or quality issues while the item is being developed, but the author or assessment expert remains in control of the final decision.

We also believe AI economics need to be transparent. Customers should not discover after adoption that successful use creates an uncontrolled consumption bill. Our preference is therefore to package AI functionality with defined usage assumptions, predictable pricing and appropriate limits rather than simply passing through variable model costs.

Finally, we apply AI selectively. Human expertise remains particularly important where validity, fairness, linguistic equivalence or high-stakes decisions are involved. AI should reduce repetitive work and focus professional effort where judgement adds the most value.

The buyer’s follow-up: Ask the supplier to model how the total price changes under low, expected and unexpectedly high AI usage. Then ask whether its commercial model rewards useful outcomes or merely increasing activity.


4. Transformation, portability and switching

Why buyers should ask this

Customers do not always remain with the same provider. A relationship may end because needs change, organisations merge, strategies move on or a different solution becomes appropriate. Losing a customer is difficult, but departure is not necessarily evidence of failure.

Both responses make positive commitments about reversing the onboarding process, providing structured exports and supporting transitions. These are welcome principles. There is also real value in suppliers showing how they have helped former customers leave successfully and professionally.

The most important practical lesson is for assessment organisations: ask the exit question at the beginning. Portability, data formats, documentation, responsibilities, retention, deletion, timescales and transition support should be understood before signing the contract, not discovered when the relationship is already ending.

The question

What do you do to reduce the cost, risk and effort of moving onto, and importantly, moving away from your products or services? Please describe how you support data portability, interoperability and transition to another supplier.

Enovate response

We believe customers should choose to remain with a supplier because the service creates value, not because leaving is prohibitively difficult. We therefore design integrations around APIs and structured data rather than closed workflows wherever possible.

Candidate, assessment, response, result and administrative data can be made available in agreed machine-readable formats, and integrations can be documented so that they can be replaced or migrated.

When onboarding a customer, we normally treat migration as a structured process: mapping existing data, validating transformations and agreeing which systems remain authoritative. The same principles apply in reverse. A transition away should include an agreed export, documentation of relevant structures and interfaces, support for validation by the incoming supplier, and controlled retention or deletion of remaining data.

This is particularly important for long-lived assessment programmes, where content and historical result data may outlive several generations of technology.

PSI response

Onboarding

We support onboarding with defined templates and a clear checklist of what's needed at each stage, so clients know upfront what to expect and what to provide.

When ACRP moved to PSI as their testing partner, the full transition with six exams and a new delivery infrastructure was completed in under 90 days. This included single sign-on from ACRP's website into the PSI Test Taker Portal, a results API for automated score reporting, and delivery through both test centres and online proctoring.

Offboarding

The same discipline applies in reverse. Clients moving to a different supplier, or bringing services in-house, get the same kind of structured template: what data will be provided, in what format, and on what timeline, so the transition is as manageable as onboarding was.

Whatever the reason a client moves on, our aim is to make the transition straightforward and to leave the relationship on good terms.

The buyer’s follow-up: Before signing, ask the supplier to describe the exit process as clearly as the implementation process. This should include formats, costs, timescales, responsibilities and support for the incoming provider.


5. Professional judgement and assessment integrity

Why buyers should ask this

One of the strongest statements in the responses is that a security approach should be rejected when its controls are disproportionate to the assessment or introduce greater risks to validity, fairness, accessibility or candidate privacy than the security benefit they provide.

This matters because security theatre is not assessment integrity. Maximum surveillance or the greatest possible number of controls does not automatically produce a valid, fair or defensible assessment. Organisations need proportionate, evidence-led security that reflects the assessment purpose, candidate population and genuine threat model.

AI has materially changed the integrity challenge. Candidates now have abundant access to powerful tools, and some programmes will require more secure delivery, including test-centre provision. But suppliers should still be prepared to warn a customer when a proposed response is disproportionate, inaccessible or likely to damage candidate trust.

An honest partner will sometimes say no. That may mean warning that a decision could undermine validity, create unacceptable accessibility problems, or expose the organisation to reputational harm. A supplier prepared to challenge its customer, rather than steering it towards the most profitable option, can ultimately build a stronger relationship.

The question

When would you advise a customer not to use a particular assessment delivery or security approach, for example online proctoring, because you believed it was inappropriate for the assessment or presented unacceptable integrity risks? How would you help them identify a better approach?

PSI response

At PSI, we believe strong client relationships depend on open, honest conversations, even when they're difficult. That means listening closely to a programme's specific needs and being willing to flex our approach, rather than offering a single fixed solution and expecting every client to fit it.

The right solution depends on the assessment, its risks and its candidate population, not on applying the same model to every programme.

Online proctoring is a good example. It's a strong delivery option, but it depends on more than the platform itself. A programme needs enough depth in its item bank to support multiple forms, and candidates need a genuinely workable testing environment, which isn't guaranteed everywhere. When those conditions aren't met, we say so directly, rather than deliver a modality that isn't right for the programme.

Sometimes the honest answer isn't yes or no: online proctoring might suit some regions or candidate groups within a programme, and a different modality suits others better.

The same thinking applies to security more broadly. We develop a bespoke security architecture that combines layered tools and controls differently depending on a programme's actual risks, rather than applying one fixed approach the same way to every client.

Enovate response

Yes. We would advise against a security approach when the controls are disproportionate to the assessment or when they introduce risks to validity, fairness, accessibility or candidate privacy that outweigh the security benefit.

Remote proctoring is a good example. It may be appropriate for professional certification where candidates test remotely and identity and test conditions must be controlled. In other situations, unreliable connectivity, accessibility requirements, the candidate population, privacy constraints or the consequences of a security breach may make a test-centre model preferable.

We therefore start with the assessment purpose and threat model rather than with a particular technology. Possible controls include identity verification, browser restrictions, human or automated supervision, additional cameras, randomisation, test-centre delivery and post-test analysis. They should be combined according to actual risk.

This also applies to content. cApStAn’s work in multilingual high-stakes assessment demonstrates that integrity is not only about preventing cheating: linguistic equivalence and avoiding construct-irrelevant differences are equally important.

The appropriate solution is therefore evidence-based, proportionate and piloted where necessary, not simply the maximum amount of technology available.

The buyer’s follow-up: Ask when the supplier last advised a customer not to buy or use one of its proposed approaches on integrity grounds, and what alternative it recommended.


Conclusion: better procurement begins with better questions

These responses reveal meaningful areas of agreement. Both suppliers describe integrated ecosystems rather than isolated platforms, AI combined with human judgement, planned continuity, structured portability, and security matched to actual risk rather than a single prescribed technology.

The answers also reinforce why buyers must go beneath feature lists. A claim that systems integrate is not the same as a candidate experiencing a coherent journey. A security certification is not the same as a tested continuity relationship. An AI feature is not an AI-native operating model or a predictable commercial proposition. An export function is not a supported exit. More security technology is not necessarily greater integrity.

The purpose of these five questions is not to identify a winner. It is to help assessment organisations have better conversations and recognise the difference between a generic assurance and a useful answer.

Only two of the 19 invited suppliers submitted responses, and I am grateful to PSI and Enovate for doing so. I would like to see more suppliers answer open industry questions like these in future. Buyers benefit when organisations are prepared to explain their thinking publicly, give practical examples and acknowledge where the answer is more complicated than a sales claim.

The five questions buyers should carry into their next supplier conversation are therefore:

  1. What does the complete candidate journey actually feel like?
  2. How will critical assessment continue during a serious cyber incident?
  3. How will AI change both the operating model and the commercial model?
  4. What happens if we eventually need to leave?
  5. When will you be prepared to tell us no?

Those questions will not make procurement simple. They should make it better.