Repair, reuse or recycle? How to choose and where AI can help

Repair, reuse and recycling require different information and decisions. AI can support parts of the analysis, from organising documents to comparing options, but technical checks and a credible destination remain essential. An illustrative furniture example shows where to start.

When a company refurbishes an office, replaces equipment or handles a return, it must decide what to do with items it no longer uses. The assessment often starts with collection and disposal. Yet some of the value may still be in the product.

A desk can be used at another site. Equipment can be brought back into service by replacing a component. A product that cannot be restored as a whole may provide spare parts to keep others working.

The circular economy encourages us to consider these possibilities before reducing a product to its constituent materials. Turning them into a project requires information, a user and an organisation capable of connecting availability with demand.

A product’s value goes beyond its materials

A product incorporates raw materials, but also design, processing, assembly and expertise. As long as it can perform a useful function, some of that value remains available.

Recycling recovers materials through transformation. Repair and reuse can also preserve the product’s structure and function, avoiding some of the processes needed to manufacture a new one.

This is the logic shown in the technical cycles of the Ellen MacArthur Foundation’s Butterfly Diagram: maintenance, reuse and the recovery of products and components keep different levels of value in circulation compared with material recycling. Ellen MacArthur Foundation.

This distinction guides the search for alternatives. It does not, however, establish the best solution for every item: condition, safety, required performance and realistic reuse opportunities must all be assessed.

Different strategies for different situations

Maintenance takes place during use to preserve performance and prevent failures. Repair addresses a specific fault to make an item usable again.

Reuse enables a product to continue performing its function, potentially with a different user. It can take place within the same organisation or through a transfer to another.

Refurbishment involves inspection and restoration to bring an item back to a defined usable condition. It is important to specify which aspects of performance are checked and what commitments are made to the next user.

Component recovery considers the usable parts of a product separately. It requires checks on compatibility, condition and reliability: an available component is not necessarily a suitable spare part.

Recycling becomes relevant when materials are recovered rather than keeping the product or its parts in use. Its effectiveness also depends on composition, separability, contamination and the availability of suitable processes and end markets.

In practice, a single batch may require several routes.

Refurbishing an office: an example of a circular choice

Consider an illustrative example. A company needs to vacate a site and manage desks, chairs and cabinets by a fixed date.

Treating all the furniture in the same way can conceal opportunities. An initial assessment could distinguish between:

• intact items suitable for another site or a new user;
• items that can be restored with limited intervention;
• items that can provide useful components;
• items suitable for material recovery, where possible.

The inventory should include quantities, dimensions, condition, photographs, location and availability dates. This information makes it possible to check potential recipients’ interest before incurring handling costs.

Timing is critical. Finding a user when the furniture must be removed the next day is very different from being able to organise the transfer several weeks in advance.

Disassembly also matters: damaging a product during removal can rule out reuse that was previously possible. The project therefore begins with the organisational decisions that precede recovery.

Where AI can help: from scattered information to a shortlist of options

The furniture example exposes a practical bottleneck: photographs, purchase records, technical sheets and requests from other sites are rarely organised in the same way. An AI-assisted workflow could help structure this information and prepare options for review. This is a proposed use case, not a claim that a deployed system has already delivered savings.

1. Build a usable inventory. Document extraction and language models can assist in turning descriptions and technical sheets into draft records: product type, model, dimensions, quantity, location and availability. Every field should retain its source. Missing information must remain missing, rather than being filled with a plausible guess. A reviewer checks the record before it is used.

2. Support an initial assessment. With suitable images and a validated model, computer vision could help group similar items or flag visible damage for inspection. A photograph does not establish structural integrity, material composition or compliance. An apparently intact chair still requires the relevant physical checks. Where records are already consistent, ordinary filters or rules may be sufficient without AI.

3. Match availability with a real need. An assistant could compare an approved inventory with requests from other sites and produce a shortlist, explaining mismatches in dimensions, quantities or delivery dates. For components, similarity is not proof of compatibility: the responsible technician must verify model references and specifications. A potential match is not a confirmed recipient.

4. Prepare comparable scenarios. Using verified quotations and explicit assumptions, the workflow could organise repair, transfer and recycling options. Costs and totals should be calculated in a spreadsheet or another controlled calculation tool, not guessed by a language model. The decision-maker reviews the inputs, uncertainty and exceptions before approving a route.

An illustrative furniture workflow: what would actually change?

Start with desks, chairs and cabinets from one office. Give each item or homogeneous group an identifier, photographs and an inspection record. The assistant prepares a draft inventory, flags missing measurements and proposes groups for review. Facilities staff then confirm condition and which items can be considered for reuse or repair.

Next, compare the approved records with another office’s actual requirements. Intact desks might be shortlisted for transfer; chairs needing replacement parts would remain pending a repair quotation and inspection; unsuitable items would be assessed for component or material recovery. No route is approved merely because the model recommends it.

The useful output is a decision sheet for each group: proposed route, supporting evidence, missing checks, recipient confirmation, total cost, deadline and accountable person. This makes the work easier to coordinate. Whether it makes the process faster or cheaper must be tested against the existing method.

Start with one task, not full automation

For a first pilot, choose a bounded task such as extracting inventory fields or matching approved items to internal requests. Use a representative sample and compare the AI-assisted approach with the manual process. Track preparation time, correction time, missing fields, incorrect matches and the proportion actually returned to use. Include software, integration and review costs before concluding that the approach pays off.

Use only data you are authorised to process, remove unnecessary personal information from photographs and documents, and choose tools with appropriate access and data-handling controls. Keep human approval for technical suitability, safety, expenditure and final destination. If exceptions and corrections outweigh the time saved, simplify the process or stop the AI pilot.

This focus on integration is consistent with the University of Oxford’s 2026 white paper on digital enablers for remanufacturing, which identifies data gaps and fragmented digital applications as barriers. The furniture workflow above is our illustrative proposal, not a case documented by that report.

Establish demand before accumulating stock

An item that can technically be reused does not automatically have a recipient. A potential user must be able to assess it and receive it in a condition that meets their needs.

The questions are practical: what characteristics are required? In what quantities? With which checks? By when? Who pays for transport and installation?

The same applies to materials and components. Availability alone does not demonstrate a market, while an expression of interest is not a confirmed purchase or use.

For an initial project, it is worth checking the destination early and separating what has been agreed from what remains an assumption. Otherwise, items may simply move from one warehouse to another without actually returning to use.

The economic outcome depends on the entire process

The sale price of a recovered item is not the same as the net value generated by the project.

Collection, sorting, disassembly, transport, cleaning, repair, testing, storage and sales management all need to be considered. Staff time and the costs of items that do not find a destination after sorting may also be significant.

The benefit varies with the chosen route:

• internal reuse may avoid a purchase;
• a sale may generate net proceeds;
• component recovery may provide a spare part;
• maintenance may extend an item’s useful life.

The comparison must start from a credible alternative. What would the company do without the project? Buy a new item, buy a used one, or decide not to make the purchase?

If furniture is transferred to a site that did not need to buy it, it is not appropriate to automatically count an avoided purchase as a saving. Similarly, the same item cannot simultaneously generate sales revenue and an internal reuse benefit.

Environmental benefits require a separate assessment

Extending product lifetimes can reduce demand for new goods and the impacts associated with their production. This is why the European Environment Agency identifies durability, repair and reuse as important levers for the circular economy. European Environment Agency.

The benefit nevertheless depends on what actually happens: how long the product remains in use, which purchase it replaces and what activities are needed to recover it.

For energy-using equipment, the use phase also matters. Keeping an existing model and replacing it with a more efficient one have effects that must be compared over the life cycle, without assuming that either option is always preferable. EEA analysis of electronics.

Technical feasibility, economic viability and environmental outcomes are therefore connected but distinct assessments. A collection or recovery indicator alone cannot capture them all.

How to set up an initial pilot

To get started, we suggest limiting the project to one product family, one site or a recurring stream. The aim is to test the entire route, from availability to the next destination, before scaling it up.

1. Define the initial problem

Which items become unused? Why? How often? What costs or difficulties does managing them create today?

The starting point should describe an observable situation, not just a broad sustainability goal.

2. Assess the alternatives

For each comparable group of items, identify feasible options and missing information. Some routes can be ruled out on technical grounds; others require discussions with maintenance providers, suppliers or potential users.

3. Assign responsibilities and timelines

Who selects the items? Who checks their condition? Who approves the costs? Who arranges the transfer and confirms receipt?

An opportunity can stall even when a technical solution exists if these responsibilities remain unclear.

4. Measure outcomes, not just activity

Useful indicators include the proportion actually returned to use, total costs, revenue or avoided purchases, the time required and any problems encountered.

It is also important to record what does not work: rejected items, overly expensive interventions and unmet requests. This information helps determine whether to change, expand or stop the project.

Preserving value requires design and organisation

Repair, reuse and recycling are not isolated decisions to be made only at the end of an item’s life. The available options also depend on how the product was purchased, designed, maintained and documented.

An initial pilot can therefore inform future choices: favour replaceable components, request technical information, plan renewals further in advance or involve suppliers in take-back arrangements.

For a business, working on circularity means making these alternatives visible and achievable. Results should be found in products that continue to serve a purpose, materials that find a use and resources whose consumption is genuinely avoided.

Turn the method into a project, or learn how to build one