SESSION 02 / SAFE AI FOR CODES, STANDARDS & RESEARCH

You know the codes. I'll show you how to make AI find them.

Three live exercises with your real questions. Five prompting techniques you'll use on Monday. One demo that changes how you research.

You are the expert. AI is a tool that puts answers in front of you faster. I'll show you how to drive it, how to catch it when it lies, and how to verify before you act. Get on Wi-Fi. Open ChatGPT, Claude, or Perplexity. Bring your hardest research question. We'll run it.

Start the session
Welcome back
Brennan Gerle

About your guide.

I'm Brennan Gerle. I run POLR AI here in Phoenix. We work with AEC firms across the Valley on where AI actually pays off inside the business.

For estimators, project managers, and field leadership. Bid leveling, scope review, RFI triage, submittal turnaround, spec and code research, daily admin. Not the hype version. The version that shows up in fewer admin hours, faster bid turnarounds, and cleaner handoffs to the field.

For this room specifically: I am not your codes expert. You are. A 30-year mason in this room knows TMS 402 better than I ever will. What I bring is how to make AI accelerate the research you already know how to do, and how to catch it when it confidently makes things up. You are the expert. AI is the tool. I am the guide.

Phoenix
POLR AI is a Phoenix-based operational AI firm working with AEC and CRE leadership teams
AEC
Custom workflows for estimating, bid leveling, spec and code research, RFI triage, and daily admin
2 of 2
Final session of the AMC AI Education Series. Take-home pack at the end.
Today's roadmap

What we'll cover.

Eleven blocks across two and a half hours. Three live exercises with your real questions. Five techniques. One demo.

01 Where the room landed since Session 1
02 Overview: web search vs deep research, citations, useful prompts
03 Live: Stump the AI, Round 1
04 Four ways AI fails
05 Five copy-paste techniques for trustworthy answers
06 Break
07 Killer demo: NotebookLM with masonry documents loaded
08 Live: Stump the AI, Rounds 2 and 3
09 The trusted sources take-home pack
10 Q&A
11 Next steps and how to take this further
Section 01 / 11 · The room since last week

What you said in the pre-work survey.

Eleven of you filled it out. Here is what the responses said. The data shaped today.

5
of you are using AI every day since Session 1
2
tried it once and want to go further
4
have not started yet. Section 04 has an easy starting point.
8
said estimating and bidding is what you want help with most

What you said you want from today.

Estimating and bidding applications
73%
Daily admin time savings
64%
Negative prompting (excluding bad sources)
45%
Asking for and verifying sources
27%
Building safe team rules for AI
27%
The bridge
You came in wanting estimating and admin help. Today is about research. Research is the foundation under both. Every bid you sharpen. Every code question you don't waste a day on. Every admin task that didn't need 30 minutes of digging.
Section 02 / 11 · Overview and best practices

Before we stump it, here's how AI actually researches.

Today is about driving AI through research. Three live exercises, five techniques, one big demo. But first: the two modes AI uses to look things up, why Perplexity is different, how each platform handles citations, and the prompts that quietly do the most work.

Two modes of AI research.

Every major AI tool has these two modes. Pick the one that matches the question.

Mode 01
Web search
Quick lookup. AI does one or two searches, pulls back a few pages, answers in seconds.
  • Best for single questions with a clear answer
  • Returns in under a minute
  • Available in ChatGPT, Claude, Gemini, Copilot, Perplexity
  • Use when you want a fast cited answer
Mode 02
Deep research
AI agent. Runs dozens of searches, follows links, iterates, and produces a structured report with citations throughout.
  • Best for multi-angle research projects
  • Takes 2 to 30 minutes depending on the platform
  • Perplexity Deep Research: 2 to 4 minutes
  • ChatGPT Deep Research: up to 30 minutes, longest reports
  • Gemini Deep Research: 100+ pages per query, Google index advantage

Why Perplexity is different.

Every other tool starts as a chatbot and adds search as a feature you have to turn on. Perplexity starts as a search engine and adds AI on top.

It always searches. Every answer comes with numbered citations inline. You can click any citation and see the exact source. There is no "I don't have current information" mode because it's always pulling from the live web.

For code and standards research, this is usually the safer default. If you only learn one new tool from today, learn this one.

How each platform handles citations.

Citations are where AI either earns trust or loses it. Know what to look for on each platform.

Perplexity
Numbered citations inline on every claim. Source list with thumbnails at top. Click anything to see the original page.
ChatGPT
Numbered footnotes when web search is on. Source list in a side panel. Without web search: usually no citations at all.
Claude
Source cards inline with the URL, title, and a snippet. Built into the answer flow.
Gemini
Source attribution at the bottom. Has a "double-check" button that highlights which claims have verified sources.
Microsoft Copilot
Numbered citations inline in Precise mode. Source list at the bottom of the response.

Prompts that quietly do the most work.

Short add-on phrases you tack onto any prompt. Each one changes what AI does behind the scenes.

What you add to the prompt What AI does differently
"Show me your citations" Forces AI to attach a source URL to every claim. If it can't, the gap becomes visible.
"Double-check your work" AI re-evaluates its own answer. Often catches its own errors before you do.
"Think carefully" Triggers longer, more deliberate reasoning. Better on multi-step questions.
"Show me only facts, no assumptions" Reduces speculation. AI sticks to what it can document.
"If you can't find an authoritative source, say so" Prevents fabrication. AI will admit when it doesn't know.
"Quote the exact passage" No paraphrase. AI returns the actual words from the source.
"What's the date on this source?" Currency check. Catches outdated info you'd miss otherwise.
"What are you uncertain about?" AI flags its own confidence gaps. Tells you what to verify.
"What would I need to verify before acting?" AI generates a verification checklist. Built-in safety net.
"Compare across multiple sources" Triangulates instead of trusting one. Surfaces disagreements between sources.
How to use this table

Stack two or three. "Show me your citations, double-check your work, and tell me what I should verify." Each phrase adds a layer of discipline. The combo is more powerful than any single one.

Now we shut all of this off and watch AI fail honestly. Round 1.

Section 03 / 11 · Live exercise · Round 1

Let's stump the AI.

First question, pulled from your survey. Ask AI cold. Verify it. Re-ask with the right prompting. Watch what changes.

Stump mode · Settings to disable
Turn these off before round 1.
OFF
Web search
ChatGPT: globe icon unclicked. Claude: model picker, no search. Gemini: use a non-search mode.
OFF
Custom instructions / Projects
Fresh chat. No project context. No saved system prompt.
-
Default model and settings
Default model. Default temperature. Type and send.
Round 1 · From the survey
"Are rebar positioners required in the Town of Gilbert when using 12-inch CMU?"
Submitted by an attendee · Estimating impact: if positioners are required and not priced, you eat the cost.
Basic prompt Are rebar positioners required in the Town of Gilbert when using 12" CMU?
Better prompt Search the web. Cite a URL for every claim. If you can't find an authoritative source, say so. Are rebar positioners required by the Town of Gilbert building department when using 12" CMU? Look for Town of Gilbert code amendments, building department interpretations, or Arizona ROC bulletins.
01 Ask AI cold. Read the answer aloud.
02 Ask the room: is this right? Where are the gaps?
03 Show the citations or lack thereof.
04 Re-ask with the better prompt. Compare.
Stump result · Claude Haiku with web search OFF
What the lame prompt got back.

I don't have specific information about the Town of Gilbert's rebar positioner requirements for 12" CMU. Building code requirements vary by jurisdiction and are regularly updated.

To get an accurate answer, you should:

1. Check the Town of Gilbert's current building code adoption (typically based on IBC/IRC with local amendments)
2. Contact the Town of Gilbert Building Department directly
3. Review the specific code section on masonry reinforcement spacing and support

Most jurisdictions require rebar positioners or similar devices to maintain consistent rebar placement in masonry, but the specific requirements depend on the code version adopted and local amendments. The building department can tell you definitively what applies to your project.

Hedge, no answer
"I don't have specific information..." Sends you to do the research yourself.
Zero citations
No URL, no section number, no source you can verify.
Confident hallucination
"Most jurisdictions require rebar positioners" - not true. They are means-and-methods, not code.
Live result · Claude Sonnet with web search and whitelist
What the better prompt actually returned.
Bottom line
No specific Gilbert requirement exists in any source I can find.
Industry code
Neither TMS 402/602 nor the model codes require rebar positioners. TMS 602 Article 3.4.B.1 only requires reinforcement be "supported to prevent displacement caused by construction loads or by placement of grout or mortar, beyond the allowable tolerances," without specifying positioners. masonryinfo.org
Town of Gilbert
Gilbert's adopted technical codes are based on 2018 IBC/IRC with local amendments. The masonry-related amendment only addresses the timing of frame and masonry inspections. No amendment requiring rebar positioners for 12" CMU appears in that document. gilbertaz.gov
State licensing
The ROC Minimum Workmanship Standards address CMU plumb tolerances and mortar joints. Nothing about rebar positioners. roc.az.gov
Industry forums
Positioners are generally treated as a means-and-methods item, not a code requirement. They are more often a specification requirement added by engineers. eng-tips.com
Tolerance rules
TMS 602 allows ±1/2 inch in width/thickness of wall and 2 inches along wall length from the core center for vertical rebar placement. Positioners help hit that tolerance, but they are not mandated. masoncontractors.org
What AI recommended next
Call Gilbert Development Services at (480) 503-6700. No published source confirms a Gilbert-specific interpretation requiring positioners for 12" CMU, and no Arizona ROC bulletin on this topic was findable. If an inspector is verbally requiring them, ask for the code section in writing.

AI sounds like an expert, and that's the danger. Remember: you are the expert. AI is simply a tool that puts answers in front of you faster.

Section 04 / 11 · Why AI fails

Four patterns. Once you see them, you can't unsee them.

What just happened in Round 1 was not random. AI fails in predictable patterns. Name them and you spot trouble before it costs you.

Fabricated citations

AI confidently quotes a URL or a code section that doesn't exist, or exists but says nothing close to what AI claimed.
"Per Town of Gilbert Code §304.5.2 ..." (No such section exists.)

Outdated information

AI's training data has a cutoff. It doesn't know the 2024 revision. It quotes the 2018 version like it's current.
Quotes an older revision of a code as if it's current. The revision changed two years ago.

Plausible wrong numbers

Close-but-wrong temperatures, dimensions, lap lengths, or thresholds. The number sounds right because it's in the ballpark. It is not the right number.
Gives a temperature, dimension, or spacing number that sounds close to right. It isn't. You can't tell unless you check the source.

Authority drift

AI cites a real source for a claim that source doesn't actually make. The source is authoritative. The claim attached to it isn't from there.
Tells you "most jurisdictions require rebar positioners." (Round 1 showed they're not actually a code requirement.)

Why this happens. The training data vs the live web.

An AI model is built by feeding it text from the internet up to a specific date. After that date, the model has no information about what was added or changed. It will still answer with confidence. Turning on web search is what lets it pull current information from live websites.

What AI knows on its own
The training data.
  • Built from text on the internet up to a specific date.
  • After that date, no information at all.
  • Answers with confidence about anything, including topics it never saw.
  • Doesn't know about anything that updated after its cutoff. Code revisions, wage rates, local rule changes.
What AI can pull in real time
The live web.
  • Web search turned on. AI looks up current information from actual websites.
  • Returns the source page, the date, and citations you can click.
  • Catches what changed since the cutoff. Current codes, current wage rates, current local rules.
  • Only as good as the sources it finds. Tell it which sources to trust, then verify.

Current knowledge cutoffs · top models · May 2026.

Model Reliable knowledge cutoff Months behind today
Claude Opus 4.7 January 2026 ~4 months
Claude Sonnet 4.6 August 2025 ~9 months
Claude Haiku 4.5 February 2025 ~15 months
ChatGPT GPT-5.5 December 2025 ~5 months
ChatGPT GPT-5.4 August 2025 ~9 months
ChatGPT GPT-4o (older) June 2024 ~23 months
Gemini 3.1 Pro January 2025 ~16 months
What this means for masonry research

Codes change. Standards change. Wage tables change. Local rules change. Without web search turned on, AI gives you whatever version it learned during training. That could be nine months old. It could be two years old. Turn on web search and use a trusted source list and you get the version that matters today.

The fix
All four happen most often when AI is working from memory instead of searching. The fix is the same every time. Turn on web search. Ask for the source. Click the link before you act on it. Five techniques. Next section.
Section 05 / 11 · The toolkit

Five techniques for trustworthy answers.

Each one is copy-paste. Each one fixes one of the patterns from Section 03. You take them back to your shop Monday.

TECHNIQUE 01 Demand sources
FIXES fabricated citations · authority drift
Without this, AI summarizes confidently and untraceably. With it, AI either produces real URLs or tells you it can't find one. Both outcomes are useful.
Add to any prompt Cite a URL for every claim you make. If you cannot find an authoritative source, say so. Do not summarize or paraphrase from your training data.
TECHNIQUE 02 Force web search
FIXES outdated information · plausible wrong numbers
Each platform handles search differently. Know which one to reach for. For technical and code research, Perplexity is the safest default because it always searches.
ChatGPT
Click the globe icon in the prompt box, or type "search the web for..."
Numbered footnotes inline. Source list in side panel.
Claude
Type "search the web" to be explicit. Search is on by default for current questions.
Source cards inline with URL, title, and snippet.
Perplexity
Always on. No toggle. This is its whole purpose.
Numbered citations inline. Source thumbnails at top.
Gemini
Always grounded in Google Search. Use the "double-check" button to highlight verified claims.
Source attribution at bottom. Color-coded verification.
Microsoft Copilot
Switch to "Precise" mode at the top of chat.
Numbered citations inline. Source list at bottom.
Add to any prompt Search the web. Do not rely on your training data. If you find conflicting sources, show me both and tell me which is more authoritative.
TECHNIQUE 03 Negative prompting
FIXES plausible wrong numbers · authority drift
Search "where can I buy concrete block in Phoenix" and Home Depot dominates the results. They sell consumer 8x8x16 blocks. You need commercial CMU from an actual block plant. Cut Home Depot out and the real suppliers surface immediately.
Without negative prompting Where can I buy concrete block in Phoenix?
With negative prompting Where can I buy concrete block in Phoenix? Exclude: Home Depot, Lowe's, Menards, Amazon, Wayfair, Etsy, consumer DIY blogs, generic comparison sites.
Reusable exclusion list Exclude these sources from all results: Home Depot, Lowe's, Menards, Amazon, Wayfair, Etsy, Pinterest, Reddit, generic blog spam, AI-generated comparison sites, consumer DIY content.
TECHNIQUE 04 Trusted sources whitelist Save as a Project or Skill
FIXES fabricated citations · outdated information · authority drift
The inverse of negative prompting. Faster than blocking everything one at a time. Tell AI exactly where to look and it stays in bounds.
Masonry research whitelist Use ONLY sources from this list: - tms.org (The Masonry Society) - ncma.org and masonryandhardscapes.org (NCMA / CMHA TEK notes) - masoncontractors.org (MCAA) - imiweb.org (International Masonry Institute) - astm.org (ASTM standards) - codes.iccsafe.org (IBC / IRC) - osha.gov (federal safety regulations) - dol.gov/agencies/whd (Davis-Bacon wage determinations) - maricopa.gov (Maricopa County codes and air quality) - azroc.gov (Arizona Registrar of Contractors) - gilbertaz.gov, phoenix.gov, scottsdaleaz.gov (municipal codes) - Manufacturer technical documentation (Echelon, Oldcastle, Acme Brick, Glen-Gery, Boral) If you cannot find the answer in these sources, say so. Do not substitute other sources without telling me.
Save this once · Talk about it later
You don't have to paste this list every time. Save it once and it loads automatically on every new chat. Three ways to do it depending on which tool you're in:
Claude Project
In Claude. Paste the rules into the project instructions. Every chat in that project follows them.
ChatGPT Custom GPT
In ChatGPT. Build a custom GPT with the rules baked in. Share it with your team.
Claude Skill
A reusable instruction set. Loads when the question matches. The most polished version.
Build once. Reuse forever. This is the bridge between today's session and what your shop actually uses next week.
TECHNIQUE 05 Verification habit
FIXES all four patterns · The habit that makes the other four matter
AI gave you a starting point. Not a finished answer. Thirty seconds of verification is the difference between a defensible bid and a costly mistake.
01
Click the URL. Does it load? Real link?
02
Check the date. Current revision? Outdated?
03
Check the author. Code body? Manufacturer? Random blog?
04
Find the passage. Does it say what AI claimed?
If you haven't tried AI yet

Five-minute starting point. Monday morning.

Open Claude or ChatGPT. Paste this prompt. Replace the question line with one real research question from your current bid. That's it. No setup. No skill. No project. Just a paste.

Search the web. Cite a URL for every claim. If you cannot find an authoritative source, say so. Use ONLY sources from: tms.org, ncma.org, masonryandhardscapes.org, masoncontractors.org, imiweb.org, astm.org, codes.iccsafe.org, osha.gov, dol.gov, maricopa.gov, azroc.gov, gilbertaz.gov, phoenix.gov, and manufacturer technical sites. Exclude: Home Depot, Lowe's, Menards, Amazon, consumer DIY, Reddit, generic blogs. Question: [PASTE YOUR REAL QUESTION HERE]

Break.

WHEN WE COME BACK · WE HAND AI THE SOURCE OURSELVES

Section 07 / 11 · The killer demo

What if you don't trust AI to find the source, and hand it the source yourself?

NotebookLM is Google's tool for grounded answers. You upload the documents. It answers only from those documents. Every claim links to the exact passage. Built for technical research. The most defensible AI workflow you can run.

The workflow.

STEP 01
Gather sources
Pull the PDFs you trust. Code excerpts. Industry tech bulletins. Safety regulations. Manufacturer specs.
STEP 02
Create a notebook
notebooklm.google.com. Free with a Google account. Upload your sources.
STEP 03
Ask the question
Use any of the five techniques. Especially technique 1 (demand sources).
STEP 04
Click the citations
Every claim links back to the source page. Verify in real time.
STEP 05
Save and reuse
The notebook persists. Add sources as you go. Build your team library.

What we have loaded for today.

Loaded · Live demo

  • Federal wage tables · Pima County · Davis-Bacon AZ20260002 (Residential)
  • Federal wage tables · Greenlee / La Paz · Davis-Bacon AZ20260029 (Building)
  • Maricopa County Rule 310 · Fugitive Dust from Dust-Generating Operations
  • Town of Gilbert · IBC amendments · 2018 Code Amendments (Technical Codes, 2020 Edition)
  • Town of Gilbert · Chapter 10 · Full Buildings & Construction Regulations

How to ask · Technique stack

  • Cite the page number for every claim
  • Quote the relevant passage exactly
  • If you find conflicting info across sources, show me both
  • If the answer isn't in these documents, say so
  • Do not summarize what's not in the loaded sources

Demo questions · pick any of these on the day.

All five are answerable directly from the loaded documents. NotebookLM will pull the exact passages with clickable citations.

Question 01 · Easy lookup · Davis-Bacon
"What's the current Davis-Bacon prevailing wage for a bricklayer on a federal residential project in Pima County? Include the base rate, fringe benefits, and any zone pay adjustments."
NotebookLM will quote $32.29 base + $9.97 fringes from AZ20260002, plus the Phoenix-radius zone pay table. Clean single-document answer. Great opener.
Question 02 · Cross-document · Wage comparison
"Compare the bricklayer prevailing wage for a residential project in Pima County to a building project in Greenlee County. Are they different? What about the mason tender labor rate?"
Pulls from both wage determination files. Brick wage is identical. Mason tender labor is dramatically different ($12.77 in Pima vs $27.41 in Greenlee). Shows AI doing real cross-document analysis.
Question 03 · Practical · Maricopa dust rule
"For a commercial construction site in Maricopa County, what dust control requirements apply to masonry work under Rule 310? What permits or plans do I need?"
Hits the meat of the 32-page Rule 310 document. NotebookLM will pull Section 300 standards and Section 400 administrative requirements with page citations. This is the "oh, this would have taken me three hours" moment.
Question 04 · Local · Gilbert walls and inspections
"For the Town of Gilbert, what are the rules around fences and walls? When does a freestanding wall need a permit and what inspections are required? Pull from both Chapter 10 and the 2018 IBC amendments."
Cross-references both Gilbert files. The 98-page Chapter 10 has the fence/wall and inspection provisions. The IBC amendments doc covers code modifications. NotebookLM pulls from both and cites the section numbers. Real-world question masonry contractors face daily.
Question 05 · The killer · Multi-document synthesis
"I'm bidding a federal building project in the Town of Gilbert. Walk me through: the brick wage I need to pay, the dust control rules under Maricopa Rule 310, what permits and inspections Gilbert requires for masonry walls, and any IBC amendments I should know about."
The magic moment. NotebookLM cross-references all five documents and produces a structured answer with citations from each. Wage tables, dust regulations, Chapter 10 inspections, and IBC amendments all in one response. This is where the audience leans forward.

When you load the source yourself, hallucination risk drops. Citations are clickable. You verify in real time. This is the workflow most contractors should adopt for spec and code research.

Section 08 / 11 · Live exercise · Rounds 2 & 3

Now we put the techniques and the tools together.

Round 2 applies the safe research workflow. Round 3 is a platform race. Three platforms, same question. You score them.

Round 2 · Workflow applied
"Per TMS 402/602, a relief angle is required at building floors or every 30 feet. Our wall is 36 feet. Do we need one?"
Submitted by an attendee · Already cites TMS
Basic prompt Per TMS 402/602, a relief angle is required at building floors or every 30 feet. Our wall is 36 feet. Do we need one?
Better prompt Search the web. Cite a URL and section number for every claim. If you cannot find an authoritative source, say so. Use ONLY sources from: tms.org, masonryandhardscapes.org, ncma.org, codes.iccsafe.org, imiweb.org, masoncontractors.org, and manufacturer documentation. Question: Per TMS 402/602, a relief angle is required at building floors or every 30 feet. The wall in question is 36 feet tall and spans between two structural floors. Do we need one? Quote the relevant code section and link to the source.
DEFINE What is the wall configuration? Where are the structural floors?
SOURCE NotebookLM with the masonry code documents loaded. Or Perplexity with the trusted source list.
PROMPT Apply techniques 1, 2, 4. Ask for the section. Turn on web search. Trusted sources only.
SUMMARIZE AI explains in plain language. Show citations.
VERIFY Click the source. Confirm the passage. Apply judgment.
Round 3 · Platform race
"What's the current Davis-Bacon prevailing wage for masonry on a federal project in Maricopa County?"
Submitted by an attendee · Cross-platform race · Estimating stakes
Basic prompt What's the current Davis-Bacon prevailing wage for masonry on a federal project in Maricopa County?
Better prompt Search the web. Cite a direct URL to the official source for every claim. If you cannot find an authoritative source, say so. Use ONLY sources from: sam.gov, dol.gov/agencies/whd, beta.sam.gov, and the official Department of Labor wage determination database. Do not summarize from third-party sites. Question: Find the current Davis-Bacon prevailing wage for masonry (bricklayer / cement mason) on a federal construction project in Maricopa County, Arizona. Quote the wage determination number, effective date, and base rate plus fringe.

Then run the better prompt across three platforms · room scores

PLATFORM A
Perplexity
Always searches. Citations top of response. Likely cleanest answer for this kind of query.
PLATFORM B
Claude with web search
Strong on synthesis. Will explain the wage determination structure as it answers.
PLATFORM C
NotebookLM
DOL PDF preloaded. Will quote the document directly. Most defensible answer.

Score the three responses. Technique matters more than tool. Picking the right tool saves time.

Section 09 / 11 · The take-home pack

Your trusted sources pack.

Print it. Save it. Tape it next to your monitor. Combine with the prompt template and you have a daily-use research workflow.

Codes & Standards

Industry associations

Federal & labor

Arizona & municipal

Manufacturers (commercial masonry)

Spec databases

Avoid for spec work

  • Home Depot
  • Lowe's
  • Menards
  • Amazon
  • Wayfair
  • Etsy
  • Pinterest
  • Reddit (most subreddits)
  • AI-generated comparison sites
  • Generic blog spam
Where this goes next

Today you typed this list into a chat. Next step: save it as a template. After that: build it into a Project or Custom GPT so it loads automatically. After that: a Claude skill or NotebookLM library that knows your sources by default. The QR code at the end takes you to POLR if you want help wiring this into your team's workflow.

Section 10 / 11 · Open the floor

Questions you've been holding.

Anything from today. Anything from Session 1. Anything you've been holding. If we run out of time, take it to polrai.com/start.

Start here
Pick one thing from today you want to try Monday morning. Bring it up and we'll workshop the prompt together.
Topics on the table
Where to start in your shop. How much paid plans cost. What to avoid. How to bring this to your team.
Off-topic welcome
Session 1 follow-ups. CRIT questions. Anything you've been chewing on since last week.
Two ways to keep going

Today is a start.

You came in wanting estimating and admin help. Today gave you the research foundation that makes both faster. If you want this dialed into your shop, pick one.

Scan to book · polrai.com/start
polrai.com bg@polrai.com