Aquaculture Procurement for Humans and AI Agents: From Production Target to a Procurement-Ready RFQ

The distinction matters because the equipment is downstream of the biology. Two farms producing the same tonnage of the same species can require completely different pumping, filtration and energy packages if one sits on a warm coastal aquifer and the other on a cold inland borehole.

Treating the project as an equipment purchase is the most common structural error in aquaculture procurement. It produces quotations that look comparable and are not, because each supplier silently assumed different densities, exchange rates, temperatures and scope boundaries.

Species and production target come first because they generate the biomass profile, and the biomass profile generates almost everything else: feed load, oxygen demand, waste production, water exchange and therefore the whole hydraulic and energy design.

Water is the second gate. Source, salinity, temperature range, suspended solids, iron and hardness all determine whether raw water can be used directly or requires pre-treatment. Pre-treatment that is discovered late is one of the most common causes of budget overrun.

Site constraints — land area, elevation profile, discharge permitting, road access, grid capacity and climate extremes — decide whether a technically sound design is actually buildable at that location.

Mechanical filtration removes uneaten feed and faeces before they break down. Biological filtration hosts the bacterial community that converts ammonia to nitrite and nitrite to nitrate. Degassing strips carbon dioxide and nitrogen, and oxygenation restores dissolved oxygen to the level the standing biomass requires.

Because the water is reused, a RAS is a tightly coupled system: a change in feed load propagates immediately into ammonia production, oxygen demand, carbon dioxide accumulation and pumping energy. This coupling is exactly why RAS specification cannot be lifted from a catalogue.

RAS reduces water consumption and improves environmental control and biosecurity. It also increases dependence on power, instrumentation and operator competence. Both sides of that trade belong in the procurement decision.

Cages exploit an existing water body and typically carry the lowest capital cost per tonne, but they expose the crop to weather, water quality variation, predation and site licensing constraints, and they offer the least environmental control.

Ponds suit warm climates, extensive or semi-intensive production and species tolerant of variable conditions. Land area and water availability are the limiting factors, and control over temperature and water quality is partial.

RAS offers the highest environmental control, biosecurity and siting flexibility, including production close to the market. It demands the highest capital investment, the highest energy input per tonne, redundant systems and skilled operation.

Hybrid configurations — hatchery and nursery in RAS, grow-out in ponds or cages — are common precisely because the systems have different strengths at different life stages.

The practical chain is: harvest weight and growth curve give cycle length; cycle length and available units give batches per year; batches, stocking numbers and survival give harvested biomass; and standing biomass at peak sets the engineering load the system must actually carry.

Peak load, not average load, sizes the equipment. A system designed around average biomass will fail in the last weeks before harvest, which is precisely when the stock is most valuable.

Safe density depends on species, life stage, water quality, temperature, oxygen supply and system type. Published universal density figures are unreliable across sites, and FishMatch does not issue generic stocking-density recommendations — those assumptions should be set and verified with the project's own engineering and biology input.

The list is stable; the sizing is not. Every item on it is derived from the same biological load, which is why a single change in production target ripples through the entire equipment schedule.

Scope boundaries matter as much as the equipment itself. Buildings, civil works, site utilities, interconnecting pipework, installation, commissioning and spares are the items most frequently excluded from an attractive-looking quotation.

Salinity source and stability shape the entire site decision: coastal intake, brackish borehole or artificial salinity each carry different pre-treatment, permitting and cost profiles.

Aeration is central. In intensive shrimp production, oxygen transfer and water movement — keeping solids in suspension and moving them to a collection point — are the operational backbone, and aeration is frequently the dominant electricity consumer.

Biosecurity design is a procurement subject, not an afterthought: water intake treatment, sourcing of post-larvae, zoning, disinfection points and effluent handling all appear in the equipment scope. FishMatch does not provide veterinary or disease-treatment advice; those decisions belong with qualified specialists.

System choice ranges from lined ponds with aeration through biofloc and greenhouse-covered systems to full indoor recirculation. Climate, land, energy price and market premium usually decide the answer, not technology preference.

Temperature control deserves particular attention. In temperate climates, heating or cooling the recirculated water can dominate both the capital package and the operating budget, and it is often quoted separately or omitted entirely.

Redundancy is the second silent variable. Backup power, duty-standby pumps, emergency oxygen and alarm systems raise the capital figure and are the reason a farm survives a failure. Comparing a redundant design against a non-redundant one on price alone is not a comparison.

A low equipment quotation can become an expensive project if buildings, utilities, redundancy, installation and operating energy are excluded from the comparison.

Price is only meaningful once scope is normalised. Two proposals differing by thirty percent frequently differ by more than that in scope, and the gap surfaces during installation.

Ask each supplier to state assumptions explicitly: design biomass, peak feed load, design temperature, oxygen supply basis, exchange or recirculation rate, energy consumption at design load, and what is excluded. A supplier quotation is only comparable when the underlying biomass, water, energy and scope assumptions are comparable.

After-sales matters disproportionately in aquaculture because the crop is alive. Response time, local presence, spare parts availability and remote support capability belong in the evaluation matrix with real weight.

Every strong manufacturer can answer these quickly. A reluctance to state assumptions in writing is itself an evaluation signal.

Request the assumptions in a structured form so that they can be compared across suppliers without re-reading twenty pages of prose per proposal.

Used well, an agent is a structuring tool. It can help a buyer articulate species, target annual production, target harvest weight, production cycles, biomass assumptions, stocking-density assumptions, mortality assumptions, water source, salinity where relevant, temperature, dissolved oxygen requirements, water exchange, recirculation rate, filtration requirements, biofiltration, solids removal, oxygenation, pumping, feeding, FCR assumptions where relevant, energy, backup systems, site constraints, climate, automation, monitoring, biosecurity, CAPEX, OPEX and project location.

Every one of those is a project input or an assumption. None of them is a fact until it has been checked against the site, the species and the engineering. Aquaculture is unusually sensitive to bad assumptions: an error in biomass, oxygen demand, water quality, stocking density, temperature or filtration capacity does not produce a slightly wrong system — it produces a system that cannot carry the crop it was bought for.

When a biological, engineering or site-specific assumption materially affects the project, uncertainty should be surfaced and reviewed — not hidden inside an AI-generated estimate.

AI agents can also assist with structured RFQ preparation and with comparing normalised supplier information. They should not be presented as approving engineering designs or selecting suppliers autonomously. On FishMatch, supported AI agents can interact with public structured content and the project's public tools; final technical assumptions and supplier decisions may still require human and engineering review.

In practice this means most projects, at least once, before the RFQ is issued. Review before issue is inexpensive. Review after ordering is a change order.

At FishMatch, project briefs are reviewed by a person before sourcing begins. Sourcing does not start automatically on submission, and suitable manufacturers are selected manually against the reviewed brief.

This distinction is deliberate. Automatic buyer-to-supplier routing produces high message volume and low relevance for both sides — a directory listing says a company exists, not that it fits this species, this scale, this climate and this budget.

Supplier discovery is useful; supplier suitability depends on the specific farm. Human procurement support remains available throughout, and buyers retain full control of the final decision.