Workshop workbook · October 2026

The AI-Powered Precast Marketer — Workbook and Free AI Toolkit

A hands-on workbook for marketing, communications and business development teams at precast concrete producers. Every prompt is ready to copy, and every tool in it has a free tier. Words in [square brackets] are yours to fill in. Sample companies, projects and specs are fictional, made up for training.

How to use this workbook

The exercises build on each other: the audit (sheet 2) feeds your content strategy doc (sheet 3), and the strategy doc feeds your website fixes, content plan, content and research (sheets 4–8). Do them in order the first time.

  • Laptop: run each exercise in ChatGPT, Claude, Gemini, Copilot or Perplexity. Use NotebookLM for long documents.
  • Phone: tap Copy on any prompt, paste it into the app, or speak it.
  • Paper: work in a team of 3–4. One person runs the prompt; the team writes and judges.
StepWhat you makeSheet
FoundationsHow AI works and how to stay accurate1
Exercise 1aHow to brief AI (R-C-T-C), then your live AI audit, scored out of 52
Exercise 1bYour content strategy doc (the foundation for everything after)3
Exercise 1cA website AI-readiness check and a fix list for your web team4
Exercise 2aA 7-day content and citation plan5
Exercise 2bOne blog, image post or carousel, reel or slide deck, with the best free tools6
Exercise 3Ideal-customer research and a list of buying signals7
Exercise 4AI agents explained, and a weekly precast news alert in your inbox8
ToolsThe free stack to do all of it9
Your planThe next 30 days10

Sheet 1 · Foundations

AI is an information power tool: it multiplies a skilled person’s output, and it does damage when handled carelessly. Two ideas cover most of what you need.

The anatomy of AI

PartBody metaphorWhat it means for you
Large language model (LLM)The brainA pattern engine that predicts the next word. It sounds confident even when wrong.
TokensSyllables and oxygenChunks of about ¾ of a word. Every page you upload uses them.
Context windowShort-term memoryHow many tokens it can hold at once. Overfill it and it forgets the start.
Generative AIVoice and handsTurns the logic into drafts: text, images, scripts.
AI agentsNervous system and limbsTakes multi-step actions with tools, such as search, files and email drafts.

Working with long documents

  • Upload the file rather than pasting 200 pages; ask for clause or page references.
  • Work section by section on very long specs.
  • When a chat drifts or repeats itself, start a fresh one.
  • Save context you reuse in a Project, Gem or custom GPT so you never retype it.

Five rules of the road

  1. Data: no confidential bids, client drawings or personal data in personal or free accounts. Use your company’s approved plan.
  2. Accuracy: an engineer signs off every technical number before it’s published.
  3. Outreach: follow CASL (Canada’s Anti-Spam Legislation): consent, clear sender ID, unsubscribe.
  4. Honesty: never pass off AI images as real projects.
  5. People decide: AI drafts and alerts; people send, bid and commit.

AI defaults to American answers. Always name the country and the standards: CSA A23.3, CSA A23.4, CPCQA certification, the National Building Code of Canada.

Sheet 2 · Exercise 1a: your live AI audit

Find out what AI tells a buyer about you today. Use a new chat in a private window so your own history doesn’t flatter the result. Keep this chat open: the content strategy doc on sheet 3 is written in the same chat.

R-C-T-C: brief AI like a smart new hire

PartWhat to giveExample
RoleWho it should bea technical marketing editor at a Canadian precast producer
ContextThe facts it can’t guessyour content strategy doc (sheet 3) and the source document
TaskOne clear jobrewrite this page as an answer-first FAQ
ConstraintsFormat, length, tone, and what it must not doa table then 5 FAQs, under 300 words; use only the source

Step 1 — the brand prompt (the buyer already knows your name):

I am considering [your company] for a precast concrete package on a [project type] in [region]. Acting as an RFP researcher for a commercial general contractor, research them and give me: portfolio and signature projects, certifications (e.g. CPCQA), strengths, weaknesses, opportunities, threats, market reputation, and the sources you used.

Score it out of 5 (1 point for each yes):

QuestionEngine 1Engine 2
Were the key brand facts correct?
Were your core product strengths accurate?
Did it mention real past projects or a positive reputation?
Did it cite your own site or trusted sources (CPCI, CPCQA listing)?
Would this answer make a buyer more likely to shortlist you?
Total out of 5

0–2: invisible · 3: visible but vague · 4–5: shortlist-ready.

Note what AI found and what it missed. Those gaps become your strategy on sheet 3.

Sheet 3 · Exercise 1b: your content strategy doc

This doc defines how your brand should be represented and optimized in AI search. Sheets 4–8 all start from it. Run the prompt in the same chat as your audit, so the AI builds on what it found. If you can’t upload documents, type your missing USPs into the prompt.

Role: You are an AEO (answer engine optimization) strategist for Canadian precast concrete producers.
Context:
- Company: [name]. Website: [URL]
- Audit results: see above (or paste the audit answer and your score)
- What AI missed or got wrong: [list the gaps]
- Our USPs and proof that AI could not find (fill in if you can't upload documents): [e.g. CPCQA categories, plant locations and capacity, 3 signature projects with year, delivery radius, awards, EPDs, fastest erection]
Task: write our Content Strategy for AI Search, with 8 sections:
1. Brand entity statement: one company name and a 25-word description to use everywhere
2. Positioning and 3 differentiators, each with proof
3. Ideal customer profile: firm types, project types, regions, project size, and trigger events that signal a new project
4. Four buyer personas (architect, structural engineer, general contractor, developer): top concerns and what convinces them
5. The 15 questions buyers ask AI that we must own, by journey stage
6. Five answer-first responses, 50 words each
7. Citation plan: 8 third-party sites where our facts must appear and match
8. Voice: tone in 3 words, and words to avoid
Constraints: use only facts from our website, the audit and the USPs above. Write [CONFIRM] for anything missing. Canadian spelling. Headings and bullets, under 900 words, ready to save as a document.

Save it as a doc. Gemini: Export to Docs. ChatGPT or Claude: download it, or copy it into Word or Google Docs. Name it “[Company] AI content strategy”. You will attach or paste it in every exercise after this one.

What is [CONFIRM]? The tag the prompt asks AI to write wherever it needs a fact it doesn’t have, instead of guessing. Each one is a fact only you can supply. Fill them in, and review the doc with sales and engineering before you publish anything from it.

Paper template (teams: fill the rows marked *)

SectionYour answer
* Brand entity statement (name + 25 words)
Three differentiators, with proof
Our USPs that AI missed
* Ideal customer: firm types, project types, regions, size
* Trigger events that signal a new project
* Architect: top concern
* Structural engineer: top concern
* General contractor: top concern
* Developer: top concern
* Five questions buyers ask AI that we must own
Third-party sites where our facts must match
Tone in 3 words; words to avoid

Sheet 4 · Exercise 1c: is your website readable by AI?

Good content is invisible if AI can’t read your site. Four checks, about 10 minutes, using your own site and Google’s free tools.

  1. Robots: open yoursite.com/robots.txt in your browser. Look for “Disallow” lines aimed at GPTBot, OAI-SearchBot, PerplexityBot, Claude-SearchBot or Googlebot.
  2. Lighthouse: run Google Lighthouse on your homepage and one product page: in Chrome, right-click → Inspect → Lighthouse → Analyze, or use PageSpeed Insights with no install. Check the SEO score and the issues it lists (blocked indexing, missing meta descriptions, crawl problems).
  3. Structured data and sitemap: put a product page into Google’s Rich Results Test and note which types exist (Organization, Product, FAQ, Article). Then open yoursite.com/sitemap.xml to confirm it exists.
  4. AI read test, then a fix list: ask your assistant: “Read [product page URL] and list every fact you can find about our products, certifications and projects.” Whatever it can’t find, a buyer’s AI can’t either. Then paste all your findings and your content strategy doc into the prompt below.
Here are the results of our website checks (robots.txt, Lighthouse, Rich Results Test, AI read test): [paste]. Using facts from our content strategy doc, write a prioritized fix list for our web developer: robots.txt changes, missing structured data with draft JSON-LD (Organization, Product, FAQPage), sitemap and page-speed issues. One plain sentence per fix explaining why it matters.
CheckWhat I foundFix needed?
Lighthouse SEO score
AI search bots allowed in robots.txt
Structured data types present
Specs published as text (not image-only PDFs)
Company facts match the strategy doc

What matters in 2026

  • Google says its AI features need no special AI file or markup: pages must be indexable and good enough to show as a snippet.
  • Let in the AI search crawlers you want (OAI-SearchBot, PerplexityBot, Claude-SearchBot). Decide separately on AI training crawlers (GPTBot, ClaudeBot, Google-Extended).
  • Add Organization, Product and FAQ structured data; publish specs and load tables as HTML text.
  • Keep company name, description and plant locations identical on your site, LinkedIn, the CPCI directory and Google Business Profile.
  • Optional: llms.txt. Cheap to add; no major engine has confirmed using it.
  • Coming: WebMCP, a proposed browser standard in a Chrome trial that lets websites offer actions to AI agents. Watch it; don’t build for it yet.

Sheet 5 · Exercise 2a: your 7-day content and citation plan

A plan that only posts on social media won’t move AI. Each day pairs one content task with one citation task, so other trusted sites start repeating your facts. Attach or paste your content strategy doc and your Deep Research report (sheet 7), update the bracketed inputs, then run:

Role: You are an AI digital marketing specialist for a Canadian precast concrete producer.
Context: our content strategy doc and our Gemini Deep Research report on ideal customers (both attached or pasted above). Before you start, use these updates:
- Ideal customer this week: [e.g. GCs bidding mid-rise residential in Metro Vancouver]
- Market focus: [region, project types]
- Key factors to promote: [e.g. new insulated wall panel line, CPCQA architectural certification, winter production capacity]
- Channels we use: [LinkedIn, website, email newsletter, trade press]
Task: build a 7-day plan that improves both our AI visibility and our pipeline, aimed at the firms and projects in the research report. Each day has one content task (image post, 30-second reel, carousel, PPT or blog) and one citation task (for example: update our CPCI member directory listing, pitch a project story to a trade publication, ask a partner GC to mention a joint project, publish a LinkedIn article answering a must-own question, update our Google Business Profile).
Table: day, goal, content task, citation task, persona, must-own question answered, owner, time needed.
Constraints: realistic for a 1-2 person team (under 90 minutes a day). Use only facts from the strategy doc, the research report and the updates above. Mark claims that need engineering sign-off [CHECK].

Save it as a doc next to your content strategy doc.

Paper teams: fill days 1–3

DayContent taskCitation taskPersonaQuestion answered
1
2
3

Sheet 6 · Exercise 2b: make the content

Pick one day from your 7-day plan and one format below. Keep your strategy doc and plan in the same chat or Project, so the AI writes from your facts. Every format follows the same rule: use only the strategy doc and your source, and mark claims needing engineering sign-off [CHECK].

Blog post

Run it in two steps: get the outline, fix the questions and any [CONFIRM] gaps, then draft in the same chat.

Step 1 · Outline

Using our strategy doc, Deep Research report and 7-day plan, outline the Day [x] blog answering "[must-own question]" for [persona]. List the 5 follow-up questions this buyer asks next (these become the H2s) and the facts from our sources that answer each. Flag missing facts [CONFIRM].

Step 2 · Draft

Now write it (about 1,200 words). H1 = the question. Directly under it, a 50-word answer that names our company, the product and the region and makes sense on its own. One H2 per follow-up question: each starts with a 1-2 sentence direct answer, then 120-180 words. Name the company, product and place; don't rely on "it" or "we". Add at least 3 facts only we have (project figures, test results, certifications) and one quote from [engineer name, title]. One spec table with a one-sentence takeaway under it. A 5-question FAQ, then a call to action. Add an author line, "Reviewed by [engineer]" and "Last updated [date]". Then give: an SEO title under 60 characters, a meta description under 150 characters, a URL slug, 3 internal-link suggestions, and Article + FAQPage JSON-LD. Plain technical tone, Canadian spelling, no hype. Use only facts from our sources. Mark claims needing engineering sign-off [CHECK].

What gets a blog cited in 2026:

  1. Answer first, in every section.
  2. Facts only you have.
  3. A named expert, and a date.
  4. Refresh every quarter.

After publishing: share it as a LinkedIn article and pitch it to trade press (citations), keep it as an HTML page rather than a PDF, use real photos with alt text, and track visits from AI assistants in Google Analytics.

Tools: Gemini (Canvas, then export to Google Docs) · ChatGPT (canvas) · Claude · Google Docs “Help me write”.

Image post or carousel

Step 1 · Copy

Write a LinkedIn image post for [persona] from Day [x] of our plan: a first line under 150 characters that hooks, 60 words total, one call to action, max 3 hashtags.

Step 2 · Image prompt for Nano Banana

Design a clean flat-vector infographic, 4:5 portrait, white background, dark text, green accents (#1E8E3E). Headline in bold sans-serif, exactly "[headline, max 8 words]". Three simple line icons labelled "[max 4 words]", "[max 4 words]", "[max 4 words]". Leave space bottom right for our logo. No people, no buildings, no other text. Also give alt text.

Real photo: upload your own jobsite photo and ask: “Keep this photo unchanged; add the headline in a white band at the top.” Carousel: make slide 1, then ask “Same style, slide 2 of 5: …”.

Make it work: max 8 words of text in the image · edit real photos, never fake a project · slide 1 first, then “same style” · always add alt text.

Tools: Nano Banana and Nano Banana Pro in the Gemini app (good at accurate text in images) · Canva · Adobe Express (Firefly) · ChatGPT images.

30-second video reel

Write a 30-second vertical (9:16) reel for [persona] from Day [x] of our plan. About 75 spoken words. Table: seconds, shot, on-screen caption (max 6 words), voiceover. Hook in the first 2 seconds, one idea, end card with [call to action]. Mark each shot FILM (real footage, with a phone shot list) or GENERATE (generic b-roll only). For each GENERATE shot, write an 8-second Veo prompt in this order: shot type, subject, one action, setting, camera movement, lighting and style, ambient sound. End each with: no on-screen text, no subtitles, no logos, no faces.

Then: film the FILM shots on a phone, generate the b-roll in Veo, and stitch them with auto-captions in Clipchamp or CapCut.

Make it work: film the real job and generate only b-roll · one Veo prompt per 8-second shot · tell Veo how the camera moves · captions on, because most people watch muted. Generated b-roll is never presented as your project.

Tools: Veo 3.1 in the Gemini app or Google Flow (generated clips) · Google Vids · Clipchamp (free in Windows, auto-captions) · CapCut.

Presentation (Google Slides)

Step 1 · Outline

From our sources, outline a 6-slide lunch-and-learn for architects on "[topic from our plan]". Slide 1 = 3 learning objectives. Each slide title is a takeaway sentence. Education, not a sales pitch. Show me the outline first.

Step 2 · Slides

Now create it. Max 25 words per slide, speaker notes of 60-90 words each, one visual suggestion per slide (real project photo or simple chart), and a final slide with 3 discussion questions. Only facts from our sources. White background, black text, green highlights.

Make it work: outline first, then slides · open with 3 learning objectives (architects often attend for continuing-education credit) · titles are takeaways, not topics · teach, don’t pitch.

Tools: Gemini Canvas (creates a deck and exports it to Google Slides, fully editable) · NotebookLM Slide Decks (a fast visual draft from your sources; the PowerPoint export is mostly images, so it’s hard to edit) · Gamma · Canva.

Sheet 7 · Exercise 3: ideal-customer research and buying signals

Step 1 — start a Deep Research report (it takes several minutes, so start it first). Open Gemini, choose Deep Research, attach your content strategy doc, and paste:

Goal: find and profile our best-fit customers for the next 12-24 months, using the ideal customer profile and buyer personas in the attached content strategy doc.
Research:
1) Planned and recently announced [project types, e.g. mid-rise residential, parkades, schools, transit] in [region].
2) The most active general contractors, architects, structural engineers and developers on those projects.
3) What these firms publicly say they value: schedule, embodied carbon, cost, quality, local supply.
4) Which structural systems and suppliers they use now.
Sources: prefer government and city websites, tender portals (BC Bid, MERX, CanadaBuys), trade press and company websites. Link every claim.
Output: a one-page summary, then a table of the 15 best-fit target firms: name, type, recent projects, why they fit our profile, and a suggested first conversation topic.

Step 2 — work with it in NotebookLM. Export the report to Google Docs. In NotebookLM, create a notebook with the report, your strategy doc and two target firms’ websites as sources. Ask: “Which 5 firms best fit our ideal customer profile, and what should we say to each?” Every answer cites the source passage. One click makes an audio overview your sales team can play on the drive to site.

Step 3 — find buying signals (Gemini or Perplexity, with web search on):

Using the ideal customer profile in our strategy doc, find buying signals in [region] from the last 90 days: rezoning approvals, building permits over $[X] million, funding announcements, design-team appointments, and tenders on BC Bid, MERX or CanadaBuys that mention precast, parkade, hollowcore or [our product]. Table: date, signal, project, owner / GC / architect, project stage, link, fit with us (1-5). Link every row.

Open at least two links before you trust the table. Spot a project at rezoning or permit stage and you can still influence whether precast is specified; by tender, the design is usually fixed. The order: rezoning → permits and funding → design team → tender → award.

Paper teams: list 5 firms or projects you know that fit your ideal customer profile.

Firm or projectSignalStageFit (1–5)

If you email anyone you find: Canada’s Anti-Spam Legislation (CASL) applies. You need consent (or a valid implied-consent basis), clear sender identification and an unsubscribe.

Sheet 8 · Exercise 4: AI agents, skills, and your weekly alert

Three terms, in plain words

TermWhat it isPrecast example
AI automationThe same steps every time, on a trigger“Every Monday, summarize new precast tenders and email me.”
AI agentYou give it a goal; it plans the steps, uses tools (browser, email, files), checks its work, and asks before anything sensitive“Research these 10 GCs and draft a first email to each for my review.”
SkillYour know-how packaged once (instructions, templates, examples) so an agent follows your playbook every time“Our brand voice”, “how we read a tender”, “our weekly signal report”

Ready-made agents you can use today for outreach drafts, research and alerts:

  • OpenClaw: open-source agent that runs on your own computer and works through chat apps. Powerful; give it only the access it needs.
  • Meta Muse: personal agent for everyday tasks such as emails, plans and bookings. US only for now.
  • Grok Bot: a team of always-on work bots from xAI that learn a workflow and repeat it.
  • ChatGPT agent and Claude Cowork: agents inside the assistants you may already use.

Whichever you use, give an agent only the access it needs, and have a person approve everything that leaves the company.

Exercise: a weekly precast news alert in your inbox

Option A — free, 2 minutes: go to Google Alerts. Create one alert per line below, then choose Show options → How often: once a week → Deliver to: your email.

  • “precast concrete” British Columbia
  • parkade Vancouver construction
  • hollowcore OR “double tee” project
  • [competitor 1] · [competitor 2]

Option B — an AI digest: create a scheduled task in ChatGPT (limits depend on your plan), or in Claude or Gemini on paid plans:

Every Monday at 8 a.m.: search the web for the last 7 days of news about precast concrete in [region / Canada]: new project announcements, tenders, rezoning approvals, and updates from [competitor 1] and [competitor 2]. Send me a 10-bullet digest with links, grouped as Projects, Competitors and Market. Flag anything that matches the ideal customer profile in our content strategy doc.

You just built your first automation. Next Monday, it reports to you.

Sheet 9 · The free AI toolkit

Every tool below has a free tier that covers this workbook. Start with one assistant plus NotebookLM; add the rest as you need them. Free limits change often, so check each tool’s current plan. List checked October 2026.

One rule for every free tool: free consumer accounts may use your chats to improve their models unless you turn that off in settings. Keep confidential bids, client drawings and personal data out of them.

AI assistants (pick one to master)

ToolFree tier gives youBest for
ChatGPTChat, web search, file upload, image generation, with daily limitsAll-round drafting, the content on sheet 6
ClaudeChat, file upload, long documents, with daily limitsLong specs, careful technical writing
GeminiChat, web search, Gems (saved assistants), limited Deep ResearchGoogle Workspace users, research reports
Microsoft CopilotChat, web search, image generationTeams already on Microsoft 365
PerplexityAnswers with cited sources, limited advanced searchesThe live AI audit (sheet 2), quick research with links

Your documents and research

ToolFree tier gives youBest for
NotebookLMNotebooks of your own files; answers only from them, with citations; audio and video overviews, mind mapsIdeal-customer research (sheet 7), your case-study library, turning a tech guide into a podcast-style briefing
Gemini Deep ResearchA few multi-page research reportsIdeal-customer research (sheet 7)

Website and AEO checks

ToolFree tier gives youBest for
Your own robots.txt and sitemap.xmlOpen yoursite.com/robots.txt and yoursite.com/sitemap.xml in a browserSeeing which AI bots you block, and whether pages are listed
Google Lighthouse (built into Chrome) and PageSpeed InsightsSEO, performance and accessibility auditsThe sheet 4 website check
Rich Results Test and Schema.org validatorShows which structured data a page hasTesting the JSON-LD from your fix list
Google Search ConsoleSearch queries, indexing problems, page reportsMaking sure Google can index your answer pages
Bing Webmaster ToolsThe same for Bing, which feeds CopilotVisibility in Microsoft’s AI search
Google AnalyticsTraffic sourcesTracking visits from chatgpt.com, perplexity.ai, copilot and gemini
Google Business ProfileYour listing in Google Maps and searchKeeping your name, address and description consistent
Any AI assistant“Read [URL] and list every fact you can find”The AI read test on sheet 4

Content creation

ToolFree tier gives youBest for
Gemini with CanvasDrafts docs and slide decks; exports to Google Docs and SlidesBlogs and lunch-and-learn decks (sheet 6)
Nano Banana and Nano Banana Pro (in the Gemini app)Image generation and editing, with accurate text in images; daily limitsImage posts, carousels and infographics
NotebookLMSlide decks, audio and video overviews from your own sourcesDecks and briefings grounded in your documents
Veo 3.1 (Gemini app, Google Flow)Short generated video clips with sound; limited on free plansGeneric b-roll for reels
Google VidsAI-assisted video editor in Google WorkspaceExplainer videos from a script
CanvaTemplates, carousels, basic AI featuresCarousels and one-pagers
Adobe ExpressTemplates, Firefly image creditsSocial graphics
GammaAI-generated slides with starter creditsFast first-draft decks
ClipchampVideo editor with auto-captions, free in WindowsEditing reels
CapCutMobile and desktop video editor with templatesVertical reels

Agents, automation and alerts

ToolFree tier gives youBest for
Google AlertsWeekly or daily email alerts for any phraseThe sheet 8 news alert
Gemini Deep ResearchMulti-page research reports with links; limited free runsIdeal-customer research (sheet 7)
ChatGPT scheduled tasksRecurring prompts that notify you; limits depend on your planA weekly AI news digest
ChatGPT agent, Claude CoworkAgents inside the assistants (mostly paid plans)Multi-step research and drafting
OpenClaw (on GitHub)Open-source, self-hosted agent framework (free; needs some technical setup)An agent that runs on your own machine

Lead signals

ToolFree tier gives youBest for
BC BidBC public tenders, saved searchesProvincial projects
CanadaBuysFederal tenders, email alertsFederal projects
MERXFree to search; some documents need a paid planPublic and private owners across Canada
City of Vancouver Open DataIssued building permitsSpotting projects before tender
Google AlertsEmail alerts for any phrase“parkade” + your region, competitor names, target GCs
LinkedInCompany pages, people search, postsFollowing architects, engineers and GCs
ApolloLimited monthly contact creditsFinding roles at target firms (follow CASL)

Industry sources to ground AI in

  • CPCI: best-practice guides, technical publications, EPDs, LCA studies, Project of the Month, Imagineering.
  • CPCQA certification: certified plants and categories.
  • CRTC guidance on CASL: the rules for commercial email in Canada.

Sheet 10 · Your 30-day plan

If you only do week 1, you’ll know more about your AI visibility than most of your competitors.

  • Week 1: review your content strategy doc (sheet 3) with sales and engineering, and fill in every [CONFIRM].
  • Week 2: send your web team the fix list from sheet 4.
  • Week 3: run your 7-day plan (sheet 5), including the citation tasks.
  • Week 4: check your weekly alert is working (sheet 8). Re-run the audit (sheet 2) and compare your score.

Questions? Connect with Vivek Sai on LinkedIn.