Glasgow-based contractor AC Whyte has worked with AI-enabled procurement specialist Erand to pilot a tool that has reduced Bill of Quantities (BoQ) processing time by 65%.
With an established presence in retrofit and cladding remediation, AC Whyte recently expanded into new build. This expansion has been accompanied by swift adoption of digital tools to support increased tender volumes, faster bid turnaround, and data-driven decision-making.
Like many contractors, AC Whyte faced a resource‑intensive tender preparation process, driven by fragmented documentation formats and the need to manually split BoQs into multiple subcontract and supplier enquiries. AC Whyte head of bid and technical development, Russell Kennedy, explained: “We competitively tender for prospective projects using a wide range of Excel, pdf, and Word documentation. The process is highly administrative, requiring tender information to be manually reviewed and separated into multiple subcontract and material enquiries.”
This manual tender‑splitting process is a high‑frequency, low‑value bottleneck that constrains bidding capacity and limits market reach. Thus, AC Whyte partnered with Erand to pilot an AI‑assisted system designed to automate repetitive tender‑packaging tasks while preserving professional oversight and accountability.
Hybrid approach
Erand was founded by CEO Sonia Piorek, a construction technology entrepreneur, and CTO Quan Han Wong, an AI expert formerly with McKinsey. Erand’s hybrid approach combines natural language processing (NLP), rule‑based trade classification, and human‑in‑the‑loop validation.
Wong revealed how the system works, starting with data ingestion and normalisation: “Raw BoQ spreadsheets with heterogeneous formats were parsed and normalised into a canonical schema. Pre-processing addressed encoding issues, line wrapping, merged cells and inconsistent field headers.
He added that a hybrid pipeline the combined:
- NLP models (tokenisation and named‑entity recognition) to identify material and labour references, as well as links to drawings and specifications;
- rule‑based ontologies mapping keywords to trade classes (eg ‘timber stud’ → ‘carpentry’); and
- heuristic scoring functions to rank probable trade assignments.
Trade packages were then generated by aggregating BoQ line items. Supporting documents (drawings and specifications) were associated using semantic similarity measures to form complete, auditable tender packages.
At this stage, the human-in-the-loop validation was introduced. “Users reviewed and corrected outputs where required,” Wong explained. “Corrections were fed back into an active learning loop to improve subsequent performance.”
Then, the system exported standardised package spreadsheets and generated pre‑populated tender emails, with optional direct email dispatch.
Performance was evaluated using:
- operational metrics (hours per BoQ and tendering capacity uplift);
- information quality metrics (precision and recall of item‑to‑trade assignment on a labelled subset); and
- usability metrics (review time and correction rates).
Test results
The pilot project involved 14 BoQs with 570 line items in each (so nearly 8,000 line items in total), plus more than 60 supporting documents and 44 trade types.
Erand’s Piorek detailed the results: “The baseline time per BoQ was circa 13 hours; with Erand that was reduced to circa 4.5 hours – a reduction of 65% in time spent packaging the BoQs.”
She highlighted other wider benefits: “Erand delivered a capacity uplift – substantially higher tender volumes could be handled without proportional staffing increases. The system also provided consistency – standardised outputs and automated document linking reduced omissions. Finally, there’s traceability – linked documentation and structured exports improved auditability.”
AC Whyte’s Kennedy noted: “Our shared objective was to significantly reduce the time spent splitting tender opportunities and to increase the volume of bids we can competitively pursue. For more complex tenders, the manual process could previously take a full day or more. With the new agent, this can now be completed in as little as 10-15 minutes, ready to be issued to prospective subcontractors and suppliers.”
Piorek concluded: “The impact of this solution extends across the contracting industry. Reducing process within bidding by three times is not an incremental improvement, it is a step-change in bidding capacity that fundamentally alters what a firm can pursue.
“Without a tool like Erand’s, scaling tender volumes requires proportional headcount increases, with the onboarding costs and coordination complexity that follow. With it, contractors can absorb a backlog of previously inaccessible opportunities and grow without first absorbing the operational drag of expanding their back office.”
