AI Agent Operational Lift for Sepi [a Division Of Transystems] in Raleigh, North Carolina

Why AI matters at this scale

Sepi, a 2001-founded civil engineering firm based in Raleigh, NC, operates in the competitive 200-500 employee band, serving public and private clients across transportation, water resources, and site development. At this scale, the firm generates a significant volume of project data—CAD files, geotechnical reports, environmental studies—but lacks the massive IT budgets of global AEC conglomerates. This creates a classic mid-market AI opportunity: enough data to train meaningful models, but a pressing need for pragmatic, high-ROI tools that don't require a team of PhDs. The shift by state DOTs toward digital delivery mandates, such as 3D model submissions, makes AI adoption not just a competitive advantage but a compliance necessity in the near future.

1. Generative Design for Accelerated Delivery

The most transformative opportunity lies in generative design for roadway and site development. Instead of manually iterating on alignments, engineers can input constraints—right-of-way limits, environmental buffers, design speed—and let AI generate hundreds of compliant alternatives. For a firm like Sepi, this compresses weeks of conceptual design into a single afternoon, allowing them to respond to RFPs with more innovative, cost-effective solutions. The ROI is direct: reduced labor hours on pursuits and a higher win rate due to demonstrating value engineering early.

2. Automated Quality Control and Digital Twin Validation

Civil engineering plans are complex, often spanning hundreds of sheets. A computer vision model trained on Sepi's historical plan sets can automatically flag missing dimensions, annotation inconsistencies, and cross-sheet clashes before the plans ever reach a senior reviewer. This reduces the expensive back-and-forth of RFIs during construction, where a single error can cost tens of thousands in change orders. Implementing this as a background check within their Autodesk or Bentley environment provides a seamless, low-friction adoption path.

3. NLP for Environmental and Regulatory Compliance

Drafting NEPA documents and permit applications is a major time sink. Fine-tuning a large language model on Sepi's archive of successful environmental assessments and categorical exclusions can produce first drafts in minutes. The model can cite precedent, suggest mitigation measures, and ensure all regulatory sections are addressed. This turns a junior engineer into a highly productive drafter, with senior staff only needing to review and stamp the final product, slashing report production costs by 40%.

Deployment Risks for the Mid-Market

The primary risk is data security and client confidentiality, especially on sensitive public infrastructure projects. Sepi must avoid feeding proprietary data into public AI models and should instead use private instances or on-premise solutions. The second risk is talent and change management; mid-career engineers may resist tools they perceive as threatening their expertise. A successful rollout requires framing AI as a 'co-pilot' that eliminates drudgery, not judgment, and investing in upskilling. Finally, professional liability insurance carriers are still evolving their stance on AI-assisted designs, so maintaining a clear human-in-the-loop for all sealed work is non-negotiable.