Introduction to AI | The Anderson Company

01 · The basics

What AI is...
and whether it's safe.

AI finds patterns in large amounts of information and turns them into a useful next step: a draft, a summary, a forecast, an answer. The question is not whether to look at it. It is what to try first, and how to do it safely.

Think of it as a very fast first-draft partner, not an authority.

01

It does not "know" like a person. It predicts a useful response from patterns, so a qualified reviewer still owns the facts, the math, and the decision.

02

It can be confidently wrong. That is fine for a first draft. It is exactly why a person checks the work before anything leaves the building.

03

Its value depends on the task. Clear inputs, a bounded ask, and a human check are what separate genuinely useful from noise.

Later, with care · agentic

This quarter · reads your files

Start tomorrow · copy & paste

"Is it safe?" is not one yes-or-no question. It depends which ring you are in. Start in the center... copy-and-paste work that never touches your network... and move outward only as trust is earned.

02 · The one we build together

The monthly cost report,
drafted in a minute.

This is the exact thing Reg raised: can AI generate the cost report, at least in draft? Yes. The real work is naming where the numbers come from. Point AI at the same sources your team already uses, ask a precise question, and let it assemble the first pass so finance spends its time on the exceptions, not the assembly.

Monthly cost report exportCost code · budget · cost-to-date · projected cost-to-complete

ERP export

Project manager updatesProgress · field conditions · pending change orders

Field notes

Change order logStatus · value · owner · aging

XLSX / PDF

The precise askDraft a one-page cost report for leadership from these sources. Explain the largest variances in plain language, separate booked costs from projections, list the open questions, and cite the source row for every number.

Monthly Cost Report

Illustrative job · leadership review

AI draft · verify

Projected revenue$28.4M

Projected margin9.8%

Open change orders$640K

Cost code	What changed	Impact
Undercut / unsuitable soils	Approved change, field-verified	+$240K
Storm & sanitary sewer	Quantities above plan	+$180K
Erosion control	Extended duration	+$96K
Asphalt paving	Buyout below allowance	−$110K

Review flag: projected profit is climbing while the booked budget stays flat. Confirm how much of the upside is recognized versus still sitting in unbooked variance before this circulates.

The win: less time assembling the first draft, more time on what actually moved. The first question AI asks back is "where does each number come from?" ... which is the conversation worth having.Illustrative figures. A real report would run on Anderson's own sources, every number traceable.

03 · Where it fits

Where could AI take work off your plate?

The best starting point is a repetitive, text-heavy task with clear inputs, a known reviewer, and a low cost of fixing the draft. A few starters by group... the real list comes from you.

01

Estimating &
Bidding

Where it could startDraft quantity takeoffs from drawings for a human to verify, compare bid documents and specs for scope and exclusions, and pull historical costs from past jobs.

Question for the roomWhere do estimators reread, reformat, or re-key the same information again and again?

02

Project
Management

Where it could startTurn daily reports and meeting notes into action logs and owner updates, draft RFIs and submittals, and surface the items still open.

Question for the roomWhat gets captured in the field but is a pain to reuse later?

03

Accounting

Where it could startDraft the monthly cost report and close narratives, code AP invoices to vendor, GL, and PO, and reduce a subcontract and its riders to plain-English net terms.

Question for the roomWhich recurring report takes more effort to explain than it does to calculate?

04

HR & Risk
Management

Where it could startAnswer policy and handbook questions in plain language, draft incident narratives and role descriptions, and track certificates of insurance and certifications.

Question for the roomWhere would a stronger first draft improve consistency without handing over judgment?

04 · Next steps

Five things to
try tomorrow.

Start small. Use an approved tool, leave sensitive information out, and treat every result as a draft until a person checks it. All five live in the safe center ring... nothing here touches your network.

1

Rewrite for clarity

Turn a dense internal note into a shorter, leadership-ready version.

2

Summarize the meeting

Convert sanitized meeting notes into decisions, owners, and next steps.

3

Stress-test a draft

Ask what is unclear, unsupported, missing, or likely to get challenged.

4

Build a checklist

Turn an approved procedure into a role-specific review checklist.

5

Explain a variance

Use fictional or approved data to draft a plain-language management narrative.

Use AI to accelerate judgment... not replace it.

A good first pilot has one owner, one defined task, a clear measure of time saved or quality gained, and an explicit human review step. Learn from it before widening the scope.

What AI is... and whether it's safe.

The monthly cost report, drafted in a minute.