When procurement teams say they do not have a technology budget for AI, they usually mean they cannot approve a new software subscription, cannot justify an enterprise platform, and cannot run a lengthy vendor evaluation process. What they typically do have is the ability to expense a $20-per-month subscription, or in many cases, they already have access to one through an existing corporate Microsoft or Google account.

That is enough. The gap is not the tool. The gap is the capability to use the tool consistently, correctly, and across the full range of tasks where AI has genuine leverage in procurement.

What "no tech budget" actually means for procurement AI

Most organisations that cannot approve a new procurement AI platform already have access to capable AI tools through existing subscriptions. Microsoft 365 Copilot is increasingly included in enterprise Microsoft licences. Google Workspace users have access to Gemini features. Teams that use these platforms may already have AI capability available that they are not using, because no one has built the workflows or the prompt library to use it effectively.

If those integrations are available, start there. The tool matters less than the workflow. A team that has built a working prompt library and knows how to structure inputs correctly will produce better results from an included Microsoft Copilot integration than a team with a premium standalone subscription and no training.

If your organisation does not have integrated AI access through existing platforms, a standalone consumer subscription is the right starting point. At approximately $20 per month per user, the cost is low enough to expense individually or to trial for a small team without a formal approval process.

The tools worth paying for at low cost

Claude and ChatGPT are both capable of handling the full range of procurement writing, analysis, and documentation tasks that generate the most time savings for procurement teams. Both have consumer subscription plans at approximately $20 per month that provide access to their most capable models. Both can handle document uploads, structured data analysis, and extended conversations with procurement context.

We use Claude in our training programmes. That is a tool choice based on our assessment of performance on procurement-specific tasks at the time of developing our curriculum. We do not recommend you treat it as definitive: both tools are updated continuously, and the relative performance on specific tasks shifts. Evaluate both with your own procurement tasks, and use the one that produces better results for your team's specific use cases.

What matters more than which tool you choose is this: do not use the free tier for serious procurement work. The free tiers of consumer AI tools have context length limitations, model access limitations, and usage caps that make them unreliable for anything beyond basic experimentation. A paid subscription gives you consistent access to the full model capability. At $20 per month, this is not a budget constraint, it is a minimum requirement for the tool to be useful.

Why the prompt library is more important than the tool

The single biggest determinant of how much value your team gets from a consumer AI subscription is not which model you are using. It is the quality and completeness of your prompt library.

A prompt library is a set of tested, structured prompts for the procurement tasks your team runs most often, RFP drafting, supplier evaluation, contract review, spend analysis, correspondence. Each prompt is designed to give the AI model the right context, in the right format, to produce a reliable, usable output in one attempt rather than three or four. The difference between a team with a working prompt library and a team without one is not subtle. One team uses AI consistently and gets consistent results. The other team uses AI sporadically, gets variable results, and reverts to manual work when the AI output requires more effort to correct than to redo.

Building a working prompt library takes time. In our training programme, teams have a baseline prompt library by the end of the second week, starting from the templates we provide and adapting them to their specific tasks. A sustainable library that covers the full range of daily procurement tasks takes approximately two months to mature. Around half of the teams we have trained continue building their prompt library independently after the programme ends.

If you are starting without training, begin with the tasks that are most repetitive and most structurally consistent: the documents you draft weekly, the correspondence you write repeatedly, the reports with a fixed format. Build one prompt at a time. Test it against real tasks. Refine the inputs until you get a reliable output. That is how the library grows.

What you can do with consumer AI tools, the full list

The range of procurement tasks where a consumer AI subscription with a working prompt library delivers genuine time savings is broader than most teams realise:

Sourcing and RFP: RFP drafting, RFP analysis and supplier comparison, shortlist rationale documentation, evaluation matrix generation, RFI questionnaire drafting.

Supplier management: Onboarding questionnaire drafting, supplier communication and chase emails, qualification summaries, supplier performance commentary, supplier news synthesis.

Contract management: Contract review and clause identification, first-draft standard contract amendments, non-disclosure agreement first drafts, contract summary and key terms extraction (note: complex PDFs are inconsistent, verify extractions against source documents).

Analytics and reporting: Spend data narrative writing, anomaly identification in structured data, management report sections, executive summaries from structured data exports, S&OP preparation documents.

Market intelligence: Supplier landscape synthesis, commodity trend summaries, regulatory change briefings, market context for category reviews.

Correspondence and documentation: Stakeholder update drafts, internal briefing documents, process documentation and SOPs, meeting preparation, negotiation preparation briefings.

Teams we have trained report that RFP generation drops from 8--15 hours to 2--3 hours after training. RFP analysis for five suppliers of medium complexity drops from 6--10 hours to approximately one hour. Those results come from consumer AI tools with a working prompt library and a trained team, not from enterprise procurement platforms.

When to invest more

There are situations where a consumer subscription is not sufficient:

When team size and volume make per-user subscription costs comparable to enterprise platform pricing, the economics of an enterprise agreement improve. Evaluate the comparison explicitly rather than assuming consumer is always cheaper at scale.

When data volume and complexity require integrations that a consumer tool cannot provide, live ERP connections, automated spend categorisation at scale, integrated contract management, the functionality requirements exceed what a consumer subscription can deliver, and platform investment is warranted.

When your organisation's data governance or security requirements prevent the use of consumer AI tools with procurement data, the tool choice is constrained by compliance, not preference. In those cases, enterprise platforms with appropriate data processing agreements are the right path.

What not to buy before you have built the baseline capability

Do not buy a procurement-specific AI platform before your team has baseline AI fluency. The platforms are built on the assumption that the team can use them. If the team does not know how to structure an AI task, how to review AI outputs critically, or how to adapt when the platform's AI features produce something unexpected, the platform investment will be underused from day one.

Do not buy because a vendor has told you that their platform's AI features will train the team. They will not. The platform gives the team a specific, constrained interface for specific, pre-defined tasks. Baseline AI fluency, the ability to use AI effectively across any task, in any tool, and to adapt as the technology changes, comes from training, not from platform access.

Build the capability before buying the platform. The capability investment pays back against any platform you subsequently choose. The platform investment without the capability behind it is a subscription to features your team will not use fully.

Frequently asked questions

Is Claude or ChatGPT better for procurement?

Both are capable of handling the full range of procurement tasks where AI has genuine leverage. We use Claude in our training programmes based on our current assessment of performance on procurement-specific writing and analysis tasks. Both tools are updated continuously, and the relative performance on specific tasks changes. The honest answer is: test both with your own procurement tasks. The tool that produces more reliable, more editable outputs for the tasks your team runs most often is the right choice for your team. The prompt library you build matters more than which tool you choose.

Can I use AI effectively without paid training?

You can get started without training. Many teams do. The typical experience is inconsistent results, sporadic use, and a gradual reversion to manual work for the tasks where AI output quality is variable. Structured training compresses the learning curve: teams that train have a working prompt library within two weeks and consistent, reliable results across a broad task range within two months. Without training, that timeline extends significantly, and some teams never get there. Training is not mandatory; it is the difference between AI being a tool your team uses occasionally and AI being infrastructure your team depends on.

What is the minimum I need to get started with AI in procurement?

A paid AI subscription, Claude Pro, ChatGPT Plus, or an equivalent, and the discipline to start building a prompt library from the first week. Begin with the three or four procurement tasks your team runs most often and most repetitively. Build a tested, reliable prompt for each one. Use those prompts consistently for a month. That is the minimum viable starting point. If the results are useful, extend the library. If they are not, the issue is the prompt structure, not the tool, and that is a trainable skill.

How do we handle data security with consumer AI tools?

This is a legitimate constraint that varies significantly by organisation. Most consumer AI subscriptions do not use inputs to train the model by default, but the data governance policies and contractual terms differ between providers and between plan types. Check your organisation's data classification policy: what category of data is your procurement information, and is that category permitted to leave organisational systems? For organisations with strict data governance requirements, enterprise AI plans with appropriate data processing agreements are the right path, even if the underlying tool is the same.

How do we build a prompt library if we are starting from zero?

Start with the tasks that are most repetitive and most structurally consistent in your team's work. Write a prompt for each one, run it against a real task, and note what needs adjusting. Refine until you get a reliable result. Save the tested version as your template. Add the next task. Do this systematically over two to four weeks, and you will have a working prompt library for your most common tasks. Our training programme provides a baseline library as a starting point, every team we train receives one, and ~50% of teams continue expanding it independently after the programme ends. Starting from zero without that baseline is slower, but the method is the same: one tested prompt at a time.

Build the capability before buying the platform.

View training options

Related reading: