SDESmith Design Engineering

Technical guidance

Computational automation and AI in building services design

Much design engineering time goes on transcription, not judgement. What happens to carbon, and to competence, if that part is automated.

Author
James Reed
Published
22 April 2026
Reviewed
29 July 2026

Key points

  • A large proportion of a design engineer’s time is spent transcribing information between calculation, drawing, schedule and specification rather than exercising engineering judgement.
  • Design option appraisal has more leverage over lifecycle carbon than any decision taken later, and it is the first activity compressed when fee and programme are constrained.
  • Automating the non-judgemental portion of MEP design therefore has an indirect but substantial carbon effect: it buys back the time in which options can actually be appraised.
  • The competence question is unavoidable — an engineer remains professionally responsible for a calculation they did not perform by hand, which requires the output to be checkable rather than merely plausible.
  • The practical test for any automated tool is whether a chartered engineer can interrogate its result and defend it. A tool that produces an unverifiable answer faster has not reduced risk; it has concentrated it.

Where the time actually goes

Ask a design engineer what they spent the week doing and the answer is rarely "engineering". It is more often re-entering the same information in four places: a calculation, a drawing, a schedule and a specification, each of which has to agree with the other three and none of which shares a source.

That transcription work is real, necessary under current tooling, and almost entirely free of engineering judgement. It is also where a large share of errors originate, because manual re-entry across four artefacts is exactly the process most likely to leave one of them out of date.

The carbon argument is indirect

The obvious environmental claim for automation — that it reduces the energy consumed by the design process — is trivially small and not worth making.

The real argument runs through option appraisal. The decisions with the greatest influence over a building's lifecycle carbon are taken early, when the design still has freedom: system selection, plant sizing philosophy, the balance of fabric against services. Appraising those options properly takes time, and it is the first thing compressed when fee and programme are tight.

If automation removes transcription work from the same fee, the time released goes somewhere. Directed at option appraisal, it has a carbon effect several orders of magnitude larger than any saving in the design process itself.

The competence problem

This is the part that cannot be waved through.

A chartered engineer carries professional responsibility for the designs they issue. That responsibility does not diminish because a calculation was produced by a tool rather than by hand, and the engineer cannot discharge it by pointing at the software.

What follows is a requirement on the tool rather than on the engineer: the output has to be interrogable. An engineer must be able to see the inputs, follow the method, identify the assumptions and form an independent view on whether the answer is right. A tool that returns a number with no accessible reasoning has not made the engineer faster; it has made them liable for something they cannot examine.

The test worth applying

For any automated design tool, the useful question is not how much time it saves. It is whether a competent engineer can take its output, reconstruct how it got there, and defend it to a peer.

Where that is possible, automation is straightforwardly good: the same judgement, applied to more options, in less time. Where it is not, the tool has concentrated risk rather than reduced work — and it has done so at exactly the point in the process where an error is least likely to be caught.

SOURCES

References

  1. CIBSE TM54: Evaluating operational energy use at the design stageCIBSE
  2. CIBSE TM65: Embodied carbon in building servicesCIBSE
  3. BS EN ISO 19650-2:2018BSI2018

RELATED

Where this comes up in practice

Services

Author

James Reed

Project Director