When Cheap Looks Credible: Protecting Quality in Evaluation and Advisory Work

In the aid and consultancy market, pressure is now coming from both sides. Budgets are being pushed down, competition is becoming harsher, and clients are increasingly expecting more to be delivered for less. At the same time, AI tools are making it easier to generate polished proposals, refined reports, and persuasive language in a fraction of the time.

That combination is reshaping the market. It is becoming easier to produce work that looks professional, confident, and complete. But faster writing does not automatically mean better thinking. The real risk is that quality will increasingly be judged by fluency and presentation rather than by evidence, judgement, and method.

Budget cuts do not just reduce costs. They reduce quality conditions.

When budgets are pushed too far, the effect is not abstract. Time for careful design narrows. Document review becomes thinner. Field engagement is scaled back. Interview coverage shrinks. Senior oversight becomes lighter. Internal challenge and quality assurance are compressed.

The final output may still look polished, but the process underneath is weaker.

That matters because evaluation and advisory work is not simply about producing text. It is about asking the right questions, selecting proportionate methods, testing claims against evidence, interpreting what findings actually mean, and being honest about uncertainty. Those are not optional extras. They are the basis of credible work.

At a certain point, low pricing stops being efficiency and starts stripping out the conditions needed for rigour.

AI lowers the cost of prose. It does not guarantee the quality of judgement.

This is the distinction the sector needs to hold onto.

AI can help with drafting, summarising, structuring material, synthesising documents, improving readability, and reducing time spent on repetitive tasks. Used well, it can support stronger workflows and free up human time for higher-value analysis.

But AI does not independently verify evidence quality. It does not know whether a sample is too thin, whether a conclusion overreaches, whether a recommendation is realistic, or whether weak data is being mistaken for a strong finding. It can help express an argument. It cannot be assumed to validate one.

That is why the current moment is so risky. As polished language becomes cheaper and easier to produce, the market may start confusing presentational quality with analytical quality. A proposal may read well without being well designed. A report may sound authoritative without being properly evidenced. A recommendation may appear decisive without being grounded in reality.

The rise of counterfeit quality

One of the clearest dangers in this environment is counterfeit quality: work that appears strong from the outside but is much weaker underneath.

Counterfeit quality is persuasive. It has clean language, clear formatting, logical structure, and fast turnaround. It creates the impression of professionalism and rigour. But appearance is not the same as substance.

This matters because procurement processes are often vulnerable to what is easiest to see. A polished bid is easier to reward than a realistic one. A fast response can look more attractive than a careful scope. A confident narrative may land better than a measured one that properly reflects constraints and uncertainty.

Over time, this can push the market in the wrong direction. Suppliers are incentivised to produce cheaper, faster, more polished outputs. Clients, often under pressure themselves, may end up rewarding responsiveness and presentation over methodological realism and evidence depth. When that happens, the less visible ingredients of quality become easier to underprice and easier to lose.

Quality cannot be judged by polish alone

In evaluation, research, and advisory work, quality is not defined by how smooth a document sounds. It is defined by whether the scope was credible, the method was appropriate, the evidence base was sufficient, the analysis was disciplined, and the limitations were handled honestly.

Good work often includes restraint. It is willing to say that the evidence is partial, that attribution is limited, that findings are mixed, or that the scope does not support a stronger claim. That may sound less impressive than a confident narrative, but it is far more valuable.

This is why polish cannot be treated as a proxy for quality. The central question is not whether an output is eloquent. It is whether it stands up to scrutiny.

Where AI can genuinely help

Rejecting AI altogether would be the wrong response. Used properly, it can be a practical and useful support tool.

It can help teams work through large document sets more efficiently, improve first drafts, clean notes, generate interview prompts, support editing, and reduce time spent on administrative formatting. These gains can be significant, especially for lean teams operating under real delivery pressure.

Used in this way, AI can create space for better work. But that only happens if the time saved is reinvested in stronger analysis, more careful review, and better judgement. If the gains are simply converted into lower fees, shorter deadlines, and higher output volume, the sector will use AI to accelerate a decline in quality rather than strengthen it.

Where AI should not replace human judgement

There are parts of evidence work that should not be casually automated away.

Scoping requires experience. Method choice requires judgement. Interviews require listening, interpretation, and contextual sensitivity. Analysis requires the ability to distinguish signal from noise, challenge assumptions, and test whether conclusions genuinely follow from the evidence. Quality assurance requires scepticism, discipline, and accountability.

These are not decorative human additions to an otherwise automated process. They are the substance of the work.

AI can assist some parts of the workflow, but it should not be treated as a substitute for professional judgement. The more the market rewards speed and low cost over substance, the greater the temptation will be to rely on tools in places where human judgement remains essential. That is where standards begin to erode.

Donors and clients need to be realistic about what credible work costs

There is also a harder message here for donors and clients.

Many organisations say they want rigorous, useful, decision-relevant evidence. But procurement structures and budget expectations often leave too little room for the very ingredients that make that possible. They may not intend to reward superficiality, but unrealistic budgets and compressed timelines can push the market towards it.

If clients want credible findings, useful recommendations, and trustworthy analysis, they need to commission work accordingly. That means setting feasible scopes, allowing proportionate time, judging bids on methodological credibility rather than polish alone, and recognising that robust work has a real cost.

Expecting high-quality evaluation or advisory support at rates that barely cover delivery is not efficiency. It is wishful thinking. In the long run, it increases the risk of poor decisions based on weak evidence presented with unwarranted confidence.

Protecting quality in the next phase of the market

The ability to produce polished text is no longer as distinctive as it once was. That is precisely why methodological integrity, evidence discipline, and honest scoping matter more now, not less.

The organisations that will retain credibility in this environment are not the ones that simply produce faster prose. They are the ones that stay clear about what AI can and cannot do, protect the substance behind the writing, and refuse to confuse polish with proof.

The question now is whether the sector will defend those standards while they are becoming less visible. Because once appearance starts replacing substance, credibility becomes much harder to recover.

The aid sector does need to adapt to tighter budgets and new technologies. But it should not do so by accepting a thinner version of quality dressed up in better language. AI can improve efficiency, but it cannot replace judgement, methodological discipline, or honest evidence standards. If donors, clients, and suppliers want credible work in this new environment, they need to be realistic about what robust analysis requires and disciplined enough to protect it.

‍