RFQ: The Last Manual Bottleneck in Industrial Sales

Strong manufacturers rarely lose opportunities because their products are weak. More often, they lose because the way they prepare quotations is too slow, too fragmented and too dependent on a handful of experts.

A modern factory can monitor equipment in real time, predict maintenance needs, optimise production schedules and trace materials across the supply chain. Yet the moment a customer sends an RFQ, many industrial companies quietly return to another era.

The request arrives by email, often as a collection of PDFs, spreadsheets, drawings and technical notes. A sales manager forwards it to engineering. An estimator copies specifications into Excel. Someone searches through old folders for a similar project. Procurement checks supplier prices, production estimates lead times, finance reviews margins, and management approves the final offer.

Days later, after several rounds of emails and corrections, the customer receives a quotation. This is a striking contradiction. Manufacturers are investing heavily in smart factories, connected equipment and automated production, while one of the most important processes responsible for generating revenue still depends on spreadsheets, email chains and the memory of a few experienced employees.

Manufacturing is becoming intelligent, but commercial execution is falling behind

The industrial sector is no longer debating whether digital transformation matters. The conversation has moved from experimentation to measurable business value.

Deloitte’s 2025 Smart Manufacturing and Operations Survey, based on responses from 600 manufacturing executives, found that 92% believed smart manufacturing would become the main driver of competitiveness over the following three years. Companies already implementing these technologies reported average improvements of 10% to 20% in production output, 7% to 20% in employee productivity and 10% to 15% in unlocked capacity. Process automation was one of the most important investment priorities identified by respondents.

Rockwell Automation’s 2026 global research shows the same shift. Of more than 1,500 manufacturers surveyed across 17 countries, 59% said they were already using smart manufacturing technologies in day-to-day operations, while only 18% remained in pilot mode. Nine out of ten described digital transformation as essential to staying competitive.

The direction is clear. Industrial companies are learning to produce more efficiently, use capacity more intelligently and make operational decisions with better data. But automation has not developed evenly across the enterprise. Factories have received sensors, production management systems, quality platforms, analytics and predictive maintenance tools. Finance has ERP. Sales has CRM. Procurement has sourcing and supplier management platforms.

The space between a customer request and a finished commercial proposal often remains fragmented. As a result, many manufacturers have digital production and analogue proposal execution.

An RFQ is not a document. It is a cross-functional decision process

From the outside, an RFQ may look like a request for a price. In a simple distribution business, that may be true. A customer specifies a standard product and quantity, and the supplier returns a price and delivery date.

Complex industrial sales operate differently. A manufacturer of electrical enclosures, automation systems, engineered components, industrial machinery or infrastructure solutions rarely receives a perfectly structured request. The customer may describe a business problem rather than a finished specification. Documents may contain contradictions, missing parameters or requirements copied from an older project. The supplier may need to recommend an architecture, select components, calculate labour, estimate production time and explain exclusions before a credible price can be offered.

The company must first understand what the customer is actually trying to achieve. It must determine which requirements are mandatory, which are negotiable and which are technically impossible. It must decide whether a standard product is suitable or whether a custom configuration is required.

Engineering then translates requirements into a design or bill of materials. Procurement checks the availability and cost of components. Operations assess workload and lead time. Finance verifies economics. Sales converts technical reasoning into a proposal the customer can understand. Management may need to approve the margin, risk exposure or non-standard contractual terms.

Where value is lost

The most visible problem is time. A complex quotation may take several days or even weeks to prepare, particularly when experts are already working on other requests. But the cost of a manual RFQ process goes far beyond labour hours.

A slow response changes the competitive dynamics of a deal. The supplier that responds first with a credible and well-structured proposal gains an opportunity to speak with the customer, clarify assumptions and influence the final decision. A technically stronger competitor may enter the discussion too late.

The second problem is dependence on individuals. In many industrial companies, proposal quality depends on a small group of senior engineers, estimators or product specialists. They know which questions to ask, which products can be combined, which configurations failed in the past and where hidden costs tend to appear.

This expertise is rarely stored as an accessible organisational capability. It lives in personal spreadsheets, email conversations, archived quotations and human memory. When those experts are busy, the quotation queue grows. When they leave, years of commercial and engineering knowledge may leave with them.

The third problem is margin leakage. Manual transfer of information between documents and systems creates countless opportunities for error. A requirement can be overlooked. An outdated component price can be used. Additional engineering work can be omitted. A delivery date can be promised without checking production capacity. A discount can be approved without understanding the full cost of the solution.

A single serious estimating mistake may cost more than the annual price of an automation platform. And when every RFQ requires many hours of senior attention, growth becomes dependent on hiring more specialists. The company cannot respond to every opportunity, so teams begin selecting which requests deserve attention. Some deals are rejected not because they are unattractive, but because the organisation does not have enough estimating or engineering capacity to prepare a response.

The business loses before the customer has even compared the proposals.

AI adoption is accelerating, but adoption is not the same as value

Artificial intelligence is spreading rapidly across corporate functions. McKinsey’s 2025 global survey found that 88% of organisations were regularly using AI in at least one business function. Yet only about one-third had begun scaling AI across the enterprise. The majority were still experimenting or running pilots.

This gap is important. Buying access to a language model is easy. Redesigning a business process around reliable, measurable and accountable AI is much harder.

High-performing organisations do not simply add AI to an existing sequence of manual tasks. They reconsider how work should move, which decisions can be automated, where human approval is required and what data the system needs to produce a trustworthy result.

RFQ automation illustrates this difference clearly. A chatbot can summarise a technical document or draft an attractive proposal. That may save a little time, but it does not solve the central problem. A manufacturer does not merely need text. It needs a technically feasible and commercially sound decision.

The system must connect customer requirements with product catalogues, engineering rules, historical quotations, pricing data, component availability, production constraints, compliance documents and approval policies.

Sales is moving towards agentic workflows

The 2026 Salesforce State of Sales report, based on a survey of 4,050 sales professionals in 22 countries, shows how quickly this transition is developing. Nine out of ten sales teams already use AI agents or expect to use them within two years. Among sales leaders whose teams already use agents, 94% consider them critical to meeting business demands. Creating quotations is already among the leading agent use cases.

This points towards a new stage of commercial automation. The first generation of enterprise AI helped employees search for information, summarise documents and generate text. The next generation will coordinate sequences of actions.

An RFQ agent may receive a request from email or CRM, identify the document type, extract technical and commercial requirements, locate missing information, compare the request with previous projects, suggest suitable products, prepare a preliminary bill of materials and send specific questions to the responsible experts.

It may then produce a first commercial draft, run compliance checks and route the proposal for approval based on deal value, discount, risk or expected margin. The human remains part of the process, but the nature of the work changes.

An engineer no longer has to start with an empty spreadsheet and search through dozens of folders. The system prepares a traceable first version. The engineer reviews assumptions, resolves exceptions and applies professional judgement.

This distinction matters. The purpose of RFQ automation is not to replace presales engineers or estimators. It is to remove repetitive mechanics so that those specialists can spend more time understanding customer problems, shaping better solutions and protecting commercial risk.

Great presales teams do not win because they complete templates faster. They win because they ask better questions and make better decisions.

What practical RFQ automation looks like

A sensible RFQ automation programme begins with intake and classification. Not every document labelled “RFQ” is actually a request for a straightforward quotation. Some documents are closer to an RFP and require solution design. Others are qualification requests, preliminary market enquiries or hybrid packages containing technical, legal and commercial stages.

The system must identify intent before selecting the workflow. It should then transform the source material into a structured requirements matrix. Instead of producing a generic summary, it should identify specifications, quantities, standards, delivery expectations, certificates, contractual obligations and commercial restrictions. It should highlight missing data and contradictions, and generate a clear set of questions for the customer.

The next stage is knowledge retrieval. The system should be able to locate comparable projects, previously approved solutions, relevant product configurations, engineering restrictions, certificates, historical cost assumptions and known risks. This turns scattered experience into an accessible organisational asset.

For standard and configurable products, the platform can go further. It can suggest an initial bill of materials, alternatives for unavailable components, estimated labour, preliminary cost and delivery assumptions. In complex engineered-to-order environments, the result will not be a final autonomous answer. It will be a strong first draft that reduces the amount of senior engineering time required.

After expert validation, the system can assemble the commercial package: the quotation, technical description, compliance matrix, assumptions, exclusions, delivery schedule and supporting documents. It can then send the package through the appropriate approval route.

Each important conclusion should remain traceable to its source. The user should be able to see which customer requirement, engineering rule, catalogue record or historical case informed the recommendation. Without traceability, faster automation may simply produce faster mistakes.

Why a large language model alone is not enough

Recent research supports both the potential and the limitations of AI in manufacturing. A 2025 systematic review of 53 research papers on large language models in digital manufacturing identified three broad areas of value: manufacturing process optimisation, data structuring and innovation, and human-machine interaction. The researchers also emphasised the importance of data quality, changing workforce roles and ethical considerations.

Another 2025 study examined the use of language models in technical services for production machinery. The researchers found that LLMs could reliably support tasks such as text correction, summarisation and question answering. Retrieval-Augmented Generation, which grounds model responses in company-specific information, improved precision and relevance. At the same time, hallucinations and integration with human workflows remained significant barriers to large-scale deployment.

Research published in 2026 on generative AI for construction bid preparation reaches a similar conclusion from another industry. Traditional bid preparation remains labour-intensive and dependent on fragmented information. Generative AI can support and partially automate the process, but the real value lies in an integrated framework rather than isolated document generation.

The lesson for industrial companies is straightforward. AI performs best when it is connected to trusted organisational knowledge, explicit business rules and a clearly defined human validation process.

Companies do not need perfect data before they begin

A common management objection is that the company is not ready. Prices are stored in ERP, customer history in CRM, product data in spreadsheets, certificates in network folders, and approvals in email. Before using AI, leaders assume they must first complete a large-scale data transformation.

Integration is certainly important. Salesforce’s 2026 Connectivity Benchmark Report found that 96% of IT leaders believe successful AI agents depend on seamless integration across systems. Half of the agents already deployed operate in isolated silos, while 86% of IT leaders are concerned that agents will create more complexity than value without proper integration.

But waiting for perfect enterprise data is rarely a realistic strategy. A manufacturer does not need to replace every core system before improving one commercial workflow. It can begin with a limited RFQ scope: one product family, one sales team or one category of customer request.

The company can identify only the information needed for that process, test the solution against historical RFQs and define clear points of human control. New data sources and product categories can be added gradually.

In this model, RFQ automation does not wait for the entire organisation to become perfectly structured. The RFQ process itself becomes a practical way to expose missing data, formalise rules and improve knowledge quality over time. Execution becomes the path to a stronger foundation.

The management question is not “Which model should we buy?”

General and commercial directors do not need to evaluate every new language model. They need to evaluate the performance of the process.

The relevant questions are operational and financial. How long does it take to move from a customer request to the first proposal draft? How many expert hours are required for one quotation? How many opportunities can the team process each month? What proportion of RFQs receive no response because capacity is limited? How often are quotations returned for correction? How much expected margin is lost through inaccurate assumptions? How does proposal speed affect win rate?

The goal is not to generate a document in seconds. The goal is to produce more accurate, competitive and commercially responsible proposals without increasing headcount at the same rate as sales volume.

That is the business case.

What will change between 2026 and 2028

Several trends are likely to reshape RFQ and proposal management over the next few years. First, automation on the buyer side and automation on the supplier side will begin to converge.

Procurement platforms are already helping buyers discover suppliers, issue requests, compare bids and manage spending. The next wave will focus more strongly on the supplier: interpreting requests, preparing technical responses, estimating solutions and managing approvals. A digital procurement process cannot reach its full potential while suppliers continue preparing responses manually.

Second, general-purpose assistants will give way to specialised industrial systems. A universal model can read a document and produce fluent text. It cannot independently understand the engineering logic behind an electrical cabinet, a customised machine, a metal fabrication project or an automation solution.

Each sector has its own standards, catalogues, cost structures, compatibility rules and production constraints. Competitive systems will therefore become increasingly vertical and domain-specific.

Third, the choice of model will become less important than the quality of the surrounding process. Companies will use multiple models and replace them as technology develops. The durable advantage will come from proprietary engineering knowledge, historical proposal data, pricing rules, integrations and learning from won and lost deals.

Finally, human review will remain essential. In industrial sales, the cost of an incorrect recommendation can be substantial. AI will prepare, check and recommend. Accountable experts will validate the result and manage exceptions.

This is not a temporary compromise. It is likely to become the standard operating model for high-value industrial decisions.

RFQ capability as a competitive advantage

A CEO does not need to understand neural network architecture to recognise an RFQ bottleneck. A few questions are enough.

How long does the company take to respond to a serious customer request? How many departments and employees participate? How much of their work is repeated from one quotation to the next? What happens when two key experts are unavailable at the same time? How many requests are ignored because the team lacks capacity? Does each proposal use knowledge from previous projects, or does the work begin again from zero?

The answers reveal whether RFQ is a controlled commercial capability or an informal chain of emails, spreadsheets and individual expertise.

At Musiakaev Lab, we see RFQ automation not as another chatbot, but as the connection of customer requirements, engineering knowledge, product configuration, pricing and approvals within one transparent process.

The manufacturers that win will not necessarily be those that deploy the largest number of AI tools. They will be those that can turn a complex customer request into a clear, accurate and economically sound proposal faster than their competitors.

Industrial companies do not buy AI for its own sake. They buy the ability to respond faster, protect margin, prepare more high-quality proposals and grow without continuously expanding the number of experts required to support every deal.