FAQs for Commercial Systems Implementation

Question 1

Is AI safe to use for contract notices?

Accepted Answer

Safe as a drafter and checker inside a controlled process, never as the process. Deadlines and legal effect stay a human's job. A contract notice is a legal act on a clock: served late, to the wrong address or with the wrong content, it can cost the whole entitlement, so the question is not whether AI writes fluent notices, it is what happens when it is wrong. Used well, AI sits inside the process: drafting a notice from a template and the facts you give it, checking a draft against the contract's requirements, summarising a chain of events into the factual background, flagging that an instruction looks like it should trigger one. Every one of those uses has a person who knows the contract reading the output before anything is sent, and the deadline itself lives in a register or diary, never in a model's memory. What is not safe is the unsupervised version: asking a general tool what notice to serve and trusting the answer, letting it decide dates, or pasting a confidential contract into a public tool against your own policies. AI also does not know your amendments: it reasons from standard forms unless you hand it the actual clause, and construction disputes are fought on amendments. Treated as a fast, tireless assistant with a supervisor, it takes real hours out of notice administration; treated as the administrator, it is an unpriced risk with a deadline attached.

Question 2

Can you set systems up without a review first?

Accepted Answer

No. The set-up is priced against the review's findings, and fixes built for problems nobody has counted end up as systems nobody uses. The review is where the price, the scope and the build order all come from, so I do not skip it, even when the problems feel obvious from the doorway. Quoting a set-up without one would mean pricing from a guess, and a guess dressed as a fixed fee is just risk moved onto whoever guessed worse. It would also mean building in an order chosen by instinct rather than by what each gap is actually costing, which is how firms end up with a beautifully configured register for a problem that was never the expensive one. If you already hold a recent and honest reading of your own systems, ask; whether it can stand in depends on whether it counted what the review counts, the hours and the exposure, rather than scoring maturity against a framework. The review's write-up is yours either way, so the worst case is knowing exactly where you stand.

Question 3

Will this need our IT department?

Accepted Answer

Rarely. The work is configuration of tools you already own, not integration, and the usual extent of IT's involvement is granting an admin permission. The build happens inside licences the business already pays for, which is what keeps IT's involvement small. There is no new infrastructure, no data migration, no integration project with a steering committee attached; the work is switching on and shaping what Microsoft 365, Google Workspace or your cost system already contains. What the build does need is access: administrator rights, or a working session with whoever holds them, so the capture steps, registers and feeds can actually be configured. In a firm with a formal IT function that is usually one conversation and a permission; in most firms it is whoever pays the subscriptions finding the password. Where something genuinely does touch IT policy, such as single sign-on, retention rules or external sharing, I say so before it is built rather than after, and that change waits for the right yes.

Question 4

Should we set up one project first or go business-wide?

Accepted Answer

Where the review does not decide it for you, pilot on one live job: a month of real running teaches more than any rollout plan. The two levels are different builds, and the review's ranked list usually makes the starting point obvious, because it orders the fixes by what each is costing. Where the call is genuinely open, the safer shape is to pilot on one live project: one job runs the new spine through a real month-end with someone beside its team, and what that month teaches gets built into the rollout instead of discovered on six jobs at once. A pilot keeps the risk on one job, gives the sceptics a working example inside their own business rather than a promise, and turns the rollout into copying something proved. The reverse also happens: a project set-up done well becomes the template the business one is built from. What matters is that the sequence is chosen from the review's findings rather than from appetite.

Question 5

What happens if the team does not use the new system?

Accepted Answer

That risk is most of the job, which is why I sit with the people through real weeks until each of them runs their part without me. A system nobody adopts is worse than no system, because everybody assumes the record is being kept while it quietly is not. So adoption is treated as the work itself rather than a training day bolted on the end. Each capture step is built to take less than a minute, inside tools the team already knows, because a step that takes longer stops happening in the first busy week. Then I sit beside the people who run each piece, through live valuations and real variations, until the awkward questions have been asked while somebody is there to answer them. And when someone routes around a step, that is read as evidence rather than disobedience: a step people avoid is usually a step built wrong, and it gets rebuilt rather than policed. Handover happens when the team runs the system without me, not when the configuration is finished.

Question 6

Will implementation mean buying new licences?

Accepted Answer

Usually not. The build starts from what you already pay for, and a new licence only appears where nothing you own honestly does the job. The build order starts in your own cupboard: Microsoft 365 or Google Workspace, the cost system, and whatever else the business already pays for, because a fix inside a known tool is cheaper, faster and likelier to stick. Most firms discover they own more capability than they use, bought in updates nobody read. Where a genuine gap exists, the position is plain: the licence is named, the reason it is needed is written down, the running cost is put in front of you, and nothing is bought until you have agreed it in writing. There are no bundled tools, no reselling and no commission anywhere in this; I take nothing from any vendor, so a recommendation to spend is only ever there because the job needs it.

FAQs for Commercial Systems Implementation, Quantity Surveying

Frequently asked questions

Before and during the build