Resource library Use case · Spend framework

Using AI to verify and improve your spend framework

For any organization that wants to manage and orchestrate its spend, two things are fundamental: access to the spend data and a framework, a classification, to associate that spend with. Data access gets all the attention. The framework rarely does.

The spend framework, also known as the spend taxonomy or hierarchy, is what lets you slice spend into meaningful categories. There's no universally right or wrong framework, as long as it gives accurate visibility, you can build and monitor strategies on it. Because of that "no wrong answer", most teams avoid investing the time to properly review an existing framework or pressure-test a new one, afraid of never-ending debates and organizational politics, especially where governance and a clear RACI are missing.

The challenge

The result, most of the time, is a poor, incomplete, vague framework that the wider organization never aligns on and therefore won't use. And when no one trusts the official framework, people quietly build their own.

Multiple frameworks,
one spend

When teams don't align on one classification, the same spend ends up categorized several different ways across the organization.

What good looks like

There are many ways to get sidetracked, but only a few elements are truly critical. A strong framework is supply-source-driven; every component is distinct and clear, including to other stakeholders; and it stays close to the transactions that actually happen, not written for an ideal world.

Supply-source driven

Built around how you actually source, so categories map to the supplier base, not to theory.

Distinct & clear

Every component is unambiguous and reads the same way to finance, operations and the rest.

Close to reality

Anchored to real transactions day in and day out, not an idealized version of the business.

We recently worked with a large manufacturing brand to revamp their procurement tools and processes. They decided to design their spend framework from scratch and partnered with Mithra to make it a successful exercise in a couple of weeks, instead of quarters. The Mithra "taxonomy builder" is purpose-built to solve exactly this for organizations of any size.

How the AI taxonomy builder works

They submitted their initial taxonomy along with spend transactions that had no associations yet. From there, the process runs in two AI-driven steps with the procurement team reviewing at every turn.

AI TAXONOMY BUILDER INPUT Initial taxonomy + transactions STEP 1 Health check flags to fix STEP 2 Proposed allocation HUMAN Review & revise OUTPUT Agreed framework iterate weekly First outlook in under a week · weeks, not quarters · 100% data-driven
Two steps, with the team in the loop. The health check flags exactly where to review, revise or define; the next step proposes a first allocation in under a week. Each weekly iteration converges the framework with reality finished in weeks, not quarters.

At the first step, the cloud solution ran a health check on the framework and pointed out precisely where the procurement team should review, revise, or write a more distinct definition.

AI Spend framework · health check
8 of 32 categories flagged
Resins & polymers · Direct materials
Clean mapping, high coverage
Clear
Agency fees · Indirect › Marketing
Overlaps with "Professional services"
Review
Freight · Logistics
No distinct definition ambiguous
Define
SaaS & cloud · Indirect › IT
Split across 3 sibling nodes
Review
Facilities · Indirect
Clean mapping, high coverage
Clear
Proposed allocation ready  91% of transactions mapped automatically
Ready for review
The health check, in the product. Categories that map cleanly are marked clear; the rest are flagged to review, revise or define, each with the reason. The team acts on evidence instead of opinion.

In the second step, algorithms tailored to this client generated the first proposed allocation of spend data against the revised taxonomy, giving the procurement team an initial outlook in less than a week. That let them review, provide feedback, and revise the taxonomy the following week.

Closing the gap with reality

On the second iteration, the power of applied AI let them see the delta between their ideal framework and what actually happens in the business day-to-day and adjust to match the two quickly. Because the whole exercise was 100% data-driven, it gave the procurement team the evidence to discuss findings and proposals with finance, marketing, and operations.

The framework was agreed upon in weeks so the team could get to the harder work of embedding it across processes and systems, confident that future changes could be simulated without slowing them down.
Weeks,
not quarters
To design and agree a framework from scratch
< 1 week
To the first proposed allocation and outlook
0%
Data-driven evidence the whole org can align on

Things to consider

A spend framework is the DNA of an organization's spend. It's unique to you, and it will keep changing with the business acquisitions, new product development, and shifts in supplier dynamics. So it needs the capability to take feedback, internal and external, simulate the change, and adjust, while remaining the single source of truth for procurement, supply chain, finance, and beyond.

Why it's worth the effort

Ignoring the framework causes fatal inefficiencies and masks key opportunities across the whole organization, not just procurement, to manage spend and to maximize innovation, service, and quality from the supplier base.

1

The framework is the DNA

Treat the taxonomy as foundational, not a formality; everything downstream depends on it being clear, distinct, and source-driven.

2

Let AI do the verification

A health check and a proposed allocation turn a politics-laden debate into a data-driven review that the whole organization can align on.

3

Keep it living

Build in the ability to take feedback, simulate, and adjust so the framework evolves with the business rather than decaying.

Building a framework from scratch, or unsure the one you have holds up? Let us run a health check on your own taxonomy and spend. For the foundations, see our guide on how to create a spend taxonomy.

Pressure-test your spend framework.

We'll run a health check on your taxonomy, propose a first allocation on a sample of your spend, and show you the delta between the framework on paper and the reality in your transactions.