Back to Sam Tomlinson
🎯
Sam Tomlinson

PPC & Google Ads

Issue #131 | How Do You Answer the “AI Strategy” Question

Sam Tomlinson <sam@samtomlinson.me>
August 31, 2025

(

link

)

***********************

Happy Sunday, Everyone!

***********************

I hope you’re taking some well-deserved time to step away from

the inbox, close the ad account, recharge and enjoy the last rays

of summer before the NFL season (Go Birds), Q4 planning and (yes,

really) 2026 arrive.

This summer has been one of the busiest in recent memory for me –

travel, events, new business, client meetings, hiring…it’s felt

like a non-stop, perpetual whirlwind. To bring some order to my

chaotic calendar, one of the things I’ve started to do is bring

my note-taker (shoutout to Otter.ai (

link

)) to every meeting AND create a process to surface insights from

those meetings.

As I’ve been going through those notes, one thing stood out: the

number of times the following question has occurred in a meeting:

“What’s our AI strategy?”

According to my (well, Otter’s) notes, it’s come up at least 16

times in the last 3 months - and that’s just virtual meetings.

It’s more than a question; it’s THE question that is

front-and-center for every CMO, CEO, investor, owner, operator

and/or marketing leader with whom I’ve had a conversation.

The problem is that it’s the wrong question.

Let’s take a step back: for the past ~2.5 years, the pace of AI

evolution in the marketing sector has been both unrelenting and

overwhelming. It feels like there’s something new - a model, an

application, a use case, a capability, a tool, a workflow -

that’s the next big thing, every single week. The sheer volume of

unsolicited messages I get from people building thinly-veiled GPT

wrappers promising staggering improvements is insane. We've

finally reached an inflection point where the noise has become so

deafening that executives across-the-board are compelled to

respond.

Just in the last few months, I’ve watched this scramble play out

seven ways to Sunday: leadership teams approving AI “pilots” just

to be seen as “doing AI.” Multiple 8/9 figure brands that have

junior employees with individual GPT licenses spinning up AI

content…with no strategy, POV or structure. And agencies are

certainly not immune – just last week, I was passed an agency

deck that had “AI” in it 19 (!!!!) times – and the worst part was

that at least half of them made no sense as a use case/workflow,

let alone something that will materially improve a brand’s P&L.

We’re in the golden age of AI theater.

Everyone wants to be seen “doing” AI (less they fall behind),

which has (rather predictably) led to organizations doing crazy

stuff – from random announcements/company-wide emails, to

assembling “AI teams” to licensing a dozen AI tools for

who-knows-what. And we’re now at the point where all that theater

has bubbled up to the leadership table, and the people that sit

around it are asking: “what’s our AI strategy?” Since we’ve all

been too busy doing AI theater, that question keeps getting

passed down until someone provides an answer that sounds both

smart and plausible.

But here’s the truth: there are very few smart answers to

incoherent questions.

AI is a capability. It’s a productivity enhancer. It’s a force

multiplier. It’s not something that sits off in a little silo

attached to your organization; it’s something with the potential

to help every constituent part of your organization perform at a

higher level.

You don’t need an “AI strategy” any more than you need an

“electricity strategy” or an “Excel strategy” – what you need is

a business strategy. AI only matters insofar as it improves your

ability to execute that business strategy more effectively and/or

efficiently.

That’s what this week’s issue is about.

Let’s start with the framing. Most companies are asking the wrong

question (“what’s our AI strategy”) – so let’s reframe it: AI is

a general-purpose technology that can either (1) reduce costs,

(2) improve choices or (3) enable entirely new capabilities.

That’s it. Every meaningful use case ladders up to one of those

three.

That changes the question to: “How do we create or compound

economic advantage using AI?”

When you approach it that way, AI initiatives start to look a lot

more like capital allocation decisions and not pet projects

designed to garner applause on business Tinder (also known as

LinkedIn). The distinction may sound small, but it matters a

great deal – because if you’re not making investment/allocation

decisions with the P&L in mind, you’re just conducting an

exceedingly expensive science fair project.

With the proper framing in place, the next question is: how do we

make AI investments that create/compound those economic

advantages? Simple: treat them like a portfolio:

* Efficiency Bets (Low Risk, Fast Payback): They’re low-risk,

capped-upside bets that reduce or eliminate the busywork/tedious

tasks that comprise 10% - 30% of your team’s time. Examples

include: workflow automations, mundane task automations, simple

customer support/success processes, research augmentation,

meeting/deliverable summarization tools. At their core, each of

these improves organizational efficiency, reduces knowledge loss

and/or improves the quality/quantity of the service you provide.

None of them are going to transform your organization – but

they’ll quickly fade into the background, quietly becoming

central to your organization in much the same way that fast WiFi

or cloud storage are critical.

* Effectiveness Bets (Medium Risk, High Impact): Every team has a

set of things they *wish* they had time to accomplish.

Effectiveness bets are - at their core - the application of AI to

the things on that list. These things focus on improving decision

quality, customer experience and/or capability

performance/execution. Think: improving your

inventory/performance forecasting, creative ideation/iteration,

developing real-time customer insights, post-click optimization,

etc. Nothing here is going to replace your marketing team, but

done well, it can make that team exponentially more effective and

efficient.

* Expansion Bets (High Risk, Asymmetric Upside): Finally, there’s

the expansion bets. This is where new categories, new AI-native

features and breakthrough products live. I would include anything

from agentic solutions to fully-automated chatbots to automated

ad managers here – these are the things we often hear hyped on

LinkedIn, but (candidly) are rarely ready for prime time. If

you’re going to make bets here, it is absolutely essential to

have both controls and continue/kill criteria in place.

This framework mirrors portfolio theory in finance: core

holdings, growth plays, and speculative bets. If this all sounds

familiar after the last two issues, it should. The principles of

compounding, risk-adjusted return and capital reallocation apply

here, too.

-----------------------------------------------------------------

The Big Takeaway: AI Theater Happens When You Confuse Tools With

Outcomes

-----------------------------------------------------------------

Flashy tools with no connection to operating leverage? That’s

theater. It’s a show that distracts you (and your team) from what

really matters.

Instead of a fancy product demo, your goal should be to craft a

compelling narrative that connects what you’re doing to how your

organization makes money. It should be straightforward, credible

and measurable. For example:

“Our primary focus with AI is improving our operating leverage.

We’ve identified 14 routine tasks that together account for 21%

of our team’s time, and we’re rolling out automated workflows for

each. We expect this to reduce the time spent on those tasks by

80%, with little-to-no impact on the quality of

deliverables/outcomes. Our secondary focus is reinvesting the

time saved above, alongside AI, into three core areas of

opportunity: (1) improving our post-click experience; (2)

creating real-time audience insights and (3) improving creative

efficiency. Based on conservative forecasts, we believe this will

allow us to reduce our customer acquisition costs by 13% in Q4

and 17% in Q1 2026. Finally, we’re also beginning several agentic

pilots that could transform the shopper experience. These are

small, strategic bets that could have a transformative impact on

the business if they hit.

That narrative is infinitely more credible than 99% of what I

hear. That’s what we should all be working toward. Bridging this

chasm requires us to re-focus what we’re doing around what

actually matters to their organization.

The reality is that most marketers aren’t lacking in intent -

they want to use AI more. The gap I see - overwhelmingly - is

infrastructure. They don’t have the tools, systems or support to

go from “I have this GPT license and a way-too-long to-do list”

to “Here’s how I can automate these 19 tasks so I can get more

done.”

Nothing that follows in this section is sexy. It’s not going to

garner the “oohs” and “ahhs” from the LinkedIn crowd. What it

will do is improve productivity and expand margins, while

allowing your team to spend more time on higher-leverage

activities.

AI tools are accelerants: integrate them in a great process, and

you’ll reap exponentially higher returns; apply them to a broken

system, and all they’ll do is multiply the chaos.

Let’s get to it:

1. Get Your Internal Data “AI-Ready”

------------------------------------

Data is the raw material that powers any AI initiative. As the

old saying goes, “garbage in, garbage out.” Most companies still

don’t know what data they have, who owns it or whether it can be

trusted. These organizations are stuck in the 2010s view of

data-as-a-warehouse, vs today’s data-as-an-operating-system.

This is not a fun shift to make. The first step is taking an

inventory of what data you have. This should include everything:

internal docs. User guides. Technical specs/docs. Processes.

Notion boards. Templates. Checklists. Content. All of it.

Odds are, you have reams of data you haven’t used in years (if

not decades) – some of which might be quite valuable, and a great

deal of which might no longer be relevant. If you just set an AI

loose on all of it, the probability is quite high that it’s going

to incorporate some of the wrong stuff in the output. Documenting

everything allows you to exclude the stuff that’s no longer

relevant, which (in turn) improves the quality of the output.

Less garbage in = more good stuff out.

This doesn’t stop there. Once you have the list, the next step is

to build taxonomies and ontologies, standardize naming

conventions, tag entities and map relationships (this process and

this process are both part of this service, etc.). The goal of

this work is to help the system understand the relationships

between all of the content/documents you’ve cataloged – so the AI

is able to grasp that Service 1 includes X, Y & Z, and is

positioned in this way to this audience and that way to that

audience.

If this sounds boring and tedious, it’s because it is. But this

is the boring, tedious work that compounds into a massive

advantage over time.

The second benefit of this is provenance. When you can prove

where your data came from, how it is secured and who touched it,

you don’t simply minimize risk - you create confidence and build

trust. That, in turn, helps you win more business.

2. Brand Taste - Clarify Your Brand, Voice & Tone

-------------------------------------------------

AI has effectively reduced the cost of creation to near-zero.

Anyone with a GPT license and some free time can create a

near-unlimited number of image, video and/or text assets. The

impact of that is that volume is no longer the differentiator;

that honor now belongs to taste (

link

).

Most brands have not codified their brand, voice or point of

view. That’s a mistake. Without guidelines, the creative outputs

(whether text, images, audio or video) are AI slop. Generic,

boring, undifferentiated sludge.

The smart move is to develop a “brand guide” that includes three

core components:

* The Non-Negotiable Rules - these are the “must follow” elements

for your brand - the things you will NEVER compromise. Your brand

promise, your desired emotional connection with your audience,

your brand values.

* The Guidelines - these are your preferences, but they’re not

written in stone. This might be the types of images you feature,

the kinds of comments/reviews you amplify, the benefits you

highlight.

* Examples/Illustrations - AI models tend to respond quite

favorably to “do this / not that” structures - so illustrate each

of the above.

The scarce resource is not content. It is attention. The question

isn’t whether you can produce at scale - anyone with a GPT

license can. The question is whether people will stop, read and

care – and that’s a question of taste.

3. Workflow Re-architecture

---------------------------

I’ve worked with hundreds of organizations, from solo ventures to

billion-dollar companies. Across every one, there are 20% of the

tasks that take 80% of the time. The exact nature of those tasks

vary, but the rule always holds.

I firmly believe the biggest gains from AI won’t come from

shaving a few minutes off a task. That kind of incremental

efficiency might garner some praise at a weekly standup, but it

doesn’t change how your business runs.

The real unlock is rethinking the task entirely. Not “how can we

do this faster?” but, “Why do we do it this way in the first

place?”

Start by mapping your top 10 recurring workflows based on volume

Ă— value. Go through your daily to-do list / hours log (if you

have one) and identify the routine tasks that take the most time

(and if you have more junior people doing the task, ask them…then

actually listen). Once identified, rebuild those workflows with

AI integrated into the system, not just sprinkled on top.

Your goal should not be to automate people away, but rather to

design processes where:

* Manual tasks are fully automated (think: copy-and-pasting data

from X to Y)

* Handoffs (i.e. marketing drafts copy, design lays it out, dev

implements) are eliminated

* Repetitive tasks (like tagging, summarizing or formatting) are

fully automated

* Human judgment is reserved for edge cases, escalation, or

strategic input

This is the difference between bolting AI onto a broken process

and using it to reduce complexity. When done right, you’re not

just speeding things up. You’re reducing the total number of

steps, decisions and dependencies – that’s what frees up your

people to do more of the work that matters.

4. Real Governance

------------------

Let’s start with this disclaimer: governance isn’t sexy. In the

move-fast-and-break-things AI world, talk of governance is

usually about as well-received as a Chris Rock skit in church.

But….no one wants to be the reason your chatbot hallucinated a

product claim or your “personalized” emails accidentally included

sensitive customer data.

That results in most companies defaulting to one of two extremes

when it comes to AI governance:

* Wild West: Everyone’s using generative tools in their

workflows, but no one knows what data is exposed, what goes into

outputs, how data is being used or what risks are accumulating

behind the scenes.

* Fortress Mode: Legal locks everything down. Teams are forbidden

from using AI tools for even basic, low-risk tasks. Innovation &

experimentation die. The company ends up watching competitors

ship, learn and iterate while they’re still drafting their usage

policy.

Both approaches fail. The first opens you up to real risk +

potential legal exposure. The second guarantees you fall behind.

The alternative - and what high-functioning, forward-looking

companies are doing - is a happy medium. That tends to look like

a lightweight governance layer that enables innovation while

still managing risk.

* Define “safe to try” vs. “needs review.” Make the criteria

explicit. If a use case uses non-sensitive inputs, generates

draft copy and isn’t customer-facing, use it! If it references

regulatory claims, pricing data or customer data, send it for

review (red light). For items that fall in between those

extremes, assign a yellow light, where someone more senior has to

weigh in before deciding on a course of action.

* Document decisions AND procedures – one of the easiest lifts

for any organization is to have a living Notion (or similar)

where decisions, procedures and prompts are stored. This allows

everyone in the company to have full transparency into what’s

being done, all while saving a ton of time (since you won’t have

to draft internal docs and your people won’t have to try to

recreate a prompt that already works).

* Make compliance a feature, not a blocker. The best guardrails

work like lane assist: they help people move faster because they

know where the edges are. I know that compliance has gotten a bad

rap over the years (and most of it is deserved), but there’s also

massive value in ensuring that everything you create/put

out/publish doesn’t pose an existential risk to your company.

Done right, governance isn’t red tape. It’s what allows teams to

test, learn and create with confidence. We’re all operating in

the digital equivalent of the wild west. Hallucinations,

copyright exposure and regulatory missteps happen every day -

what is needed is a way to minimize the risk while maintaining

the upside, which is exactly what this provides.

5. Talent & Training

--------------------

This may be obvious, but paying for a bunch of AI tools doesn't

magically make your people more effective, competent or

productive.

Training your people on how to use those AI tools does.

Your goal should be to make two fundamental changes: (1) shift

from individual instances to scalable use cases and (2) move from

hiring for technical excellence to judgement.

Let’s start with (1). Just the other day, I was talking with one

of our digital team members who was “vibe coding” a JavaScript

setup for landing pages. He built conditional redirects off a

radio button selection, sending users to different thank-you

pages based on their response to a form question. The idea was to

provide both a tailored experience AND better analytics data.

It’s a clever use case. The problem was that it was built for a

single client, when we have at least 5 that would benefit from it

being implemented. We solved that by adding this to our library

of effective prompts, then sharing it with the entire team – so

other account teams who could use it were able to access and

implement it in minutes. And as each of those client teams

implemented the workflow, they uncovered several weird

edge-cases, which we then documented and used to further refine

it.

There’s nothing magical about this – it’s just training. The

difference is that this training transforms one person’s clever

experiment into a scalable capability the entire agency can use,

and that all of our clients can benefit from.

That’s the difference between tinkering and leverage. Individual

hacks live and die inside one client account or team member’s

head. Scalable use cases are documented, shared, refined and

eventually become part of the “operating system”. When you build

that kind of library, every new engagement/project starts a step

ahead of the last one.

This brings me to the second shift (and this is one that’s going

to have HR furious): stop hiring purely for technical excellence.

Tools are closing that gap every day. What you cannot automate is

good judgment. Knowing when an output is good enough. Spotting

when an edge case will break something important. Choosing which

ideas to run with and which ones to put back in the box.

AI makes technical execution cheaper. That, in turn, pushes the

premium to synthesis, taste & judgement. The teams that hire

people with those skills will quickly find themselves at a

massive advantage, because their people will be better equipped

to manage increasingly-more-sophisticated technical tools.

6. Measure The Right Thing

--------------------------

Let’s get clear about one thing: the goal of any/all of this is

NOT “AI adoption” - it’s better business economics. That’s the

whole reason we’re doing this, so we need a

measurement/accountability system that aligns with that goal. I’d

recommend you focus on three core areas:

* Operating Leverage: are our investments in AI tools allowing us

to realize real cost and/or time savings on our operations? Are

we able to automate/expedite routine tasks, thereby freeing up

our team to do higher leverage work? Are these tools reducing the

time required to produce key deliverables (i.e. monthly reports)?

* Effectiveness: Doing something faster or cheaper is only good

IF the quality of the output is as good as (or, ideally, better

than) what was produced before. If you save 3 hours by automating

a client report (good!) but the quality of the automated report

is so poor/disjointed that your team now needs to spend 4 hours

explaining to the client/stakeholder the content and what it

means, you haven’t actually accomplished a damn thing. The same

holds true for AI-generated landers or creatives: they should be

as effective, if not more effective, than what you had before.

The goal must be not only to improve efficiency, but to improve

effectiveness.

* Visibility: This is the one most people forget. How often is

your brand showing up in AI answers? How much traffic is coming

from assistants like ChatGPT or Perplexity? Is the brand visible

in the new discovery surfaces like AI overviews or AI mode, where

customers are starting their journey? This isn’t just about

working smarter inside the walls; it’s about making sure the

brand isn’t invisible as these tools re-write how people find

things.

At the end of the day, the only thing that matters is this: is

your organization more profitable, more effective and more

discoverable than you are today? That’s the thing we’re all

trying to figure out how to do, and the best way to accomplish it

is by having a scorecard that maps to it.

------------------------------------

6 Tactical Moves You Can Execute Now

------------------------------------

If the preceding section is theoretical, this one is exceedingly

practical. The reality is that AI is here to stay, and there are

plenty of ways you can integrate it into your day-to-day right

now.

Here are a few things that we are doing and what I think

marketers should be doing yesterday:

1. Optimize for AI Overviews + Assistants

-----------------------------------------

Search isn’t dead. It’s evolving.

SparkToro’s latest research (

link

) shows that while just 20% of Americans use AI tools more than

10 times a month, that usage is enough to make a dent,

particularly among high-intent, early-adopter audiences. But

here’s the kicker: traditional search traffic hasn’t meaningfully

declined. Which means Ai tools are NOT a channel replacement;

they’re a visibility expansion opportunity (something I’ve said

for years - AI will make people search more, not less):

The takeaway: it’s not “either/or.” It’s “yes/and.” AI Overviews

are carving out top-of-funnel real estate in addition to classic

rankings. And, for now, the competition for that space is light…

which makes this a rare opportunity.

Fortunately (and despite all the AI bro comments to the

contrary), appearing in AI Overviews is not a super-secret

science; it’s just good, old-fashioned SEO. If you want to appear

in more AI Overviews, here’s the quick playbook:

* Identify which searches are impacted by AI overviews (yes, you

will have to do some searching!). For the ones that are, note

which sites are included (yes, you’ll have to do some clicking).

* Document the commonalities + the weaknesses in the AI Overview

– if an answer has wrong/outdated/partially correct/incomplete

information, that’s an opportunity!

* Once you’ve identified the opportunities, build pages designed

for conversational intent: “Best X for Y in Z context”

* Create layered content: 1-sentence takeaways, 1-paragraph

summaries, and full deep dives, all on the same page

* Ensure your content actually communicates real, legitimate,

unique value – if you’re just saying the same things as everyone

else, there’s little reason for any engine to include your

content in their outputs.

* Structure pages cleanly: H1–H3s, FAQs, comparison tables,

glossary sections

* Add schema (FAQ/Product/How-To)

* Fix page speed + crawl issues

* Publish content with real bios, citations and external

credibility signals

Remember, you’re writing for people, not keywords or language

models. If you create content that adds value to your audience

and is easily intelligible to the LLM, you’ll quickly find

yourself included.

2. Treat AI Assistants as Channels

----------------------------------

Most CMOs I’ve spoken to are still treating referrals from AI

tools (Perplexity, Claude, ChatGPT) as a curiosity, not a true

channel. That’s a massive mistake. We’ve had several clients that

have gone from 5% in the last 90 days. That may not sound huge

(and, even at 5%, it’s the 8th-largest source of traffic), but

the rate of increase is shocking.

This is the SEO goldrush all over again. The numbers are/will be

small at first, until they’re too overwhelming to ignore. The

solution is to begin tracking referrals from ChatGPT, Perplexity,

Claude, et al now. Spend time understanding the differences in

session types/quality between AI referrals and standard search

(this is a great use case for Heatmap.com (

link

) or Microsoft Clarity).

Once you have clarity on how much you’re appearing now, the next

step is to increase your visibility. Publish explainers and

glossaries in crawlable, indexable formats. Contribute to public

knowledge graphs where appropriate.

When clients ask if this is worth the effort, my answer has been:

“If your target audience is asking questions (doesn’t matter if

it’s on Reddit or to ChatGPT), you want your content to be the

answer.”

That starts by treating these assistants like the channels they

are.

3. Own the Post-Click Experience

--------------------------------

So much of the conversation around AI overviews + their impact on

search right now is on visibility (who is showing up, how much

traffic is it generating, etc.) – but none of that matters if

your post-click experience doesn’t convert. The reality is that

AIOs tend to drive longer, more conversational queries (something

we’ve known for quite some time) – which means that expectations

of your post-click experience are higher.

A generic “contact us” page isn’t going to cut it for a user who

just typed “best CRM for AdTech B2B SaaS companies in US” – that

user expects the lander to match the intent of their query.

The reality is that most brands are still playing catch-up in

this department.

Align your landing experiences with the shape and structure of

long-form, conversational queries (filter your Google Search

Console for queries with >8 words). These queries aren’t

transactional in the traditional sense; they’re layered with

context and constraints.

That means building modular, targeted landing pages that speak to

real-world situations: “best platform for X use case,” “solutions

under Y budget,” “how this compares to Z.” Your goal for each of

these pages should not just be to sell the service/product - it

should be to help your visitor make a smart, confident decision.

This approach creates space for compounding returns. Why? Because

once you find a variant that lifts conversion or AOV, whether

it’s a headline, a section order, a pricing frame or an

objection-handling block that improvement holds across every

future visitor who fits that same profile.

The process for this is simple (and can be easily augmented with

AI):

* Build modular, intent-matched landers that speak to specific

use cases, constraints, and objections

* Run A/B tests that focus on downstream metrics like CVR, AOV,

and lead quality

* Roll learnings back into your core templates so each new page

gets better by default

Then repeat. Iterate. Stack improvements.

This is how post-click becomes a flywheel. Not just a place to

recover wasted traffic but a surface where smarter decisions and

better experiences multiply over time.

4. Ask AI About Your Brand, Then Fix The Gaps

---------------------------------------------

Have you asked ChatGPT what it thinks about your company? Or

Perplexity to describe your product/service? Or Gemini to tell

you what sets your business apart from your competitors? If not,

you should. Chances are, the answers are incomplete, outdated, or

just plain wrong.

That’s a brand risk hiding in plain sight.

The defensive move is to seed the internet with clarifying

content. Publish canonical explainers. Create source-of-truth

pages that models can pull from. For regulated industries, sign

and date them. Give the models something better to reference.

If you don’t, someone else will.

There’s an offensive play here, too. Most of your competitors

haven’t done any of this, which means their AI footprint is thin,

stale or non-existent. Your goal should be to treat every

unanswered question about your competitors as an opportunity to

frame the conversation in your favor. If an assistant generates a

side-by-side comparison, the company with fresh explainers and

recently updated buyer’s guides will look credible, while the one

with a two-year-old press release will look asleep.

You should absolutely defend your own presence. And you should

fill the gaps your competitors leave. Any information vacuum will

be filled; the question is whether that’s by you or by someone

else.

5. Increase Content Velocity Without Going Generic

--------------------------------------------------

AI can (and will) generate endless creative. Without structure,

judgment and taste, most of it will be AI slop. If you want to

avoid publishing reams of generic garbage, the key is to revise

your workflow:

Step #1: Use AI to help uncover patterns in how your audience is

searching for information/solutions (customer insights)

Step #2: Based on that, have AI draft content that is likely to

appeal to your audience as they’re searching for that information

Step #3: Feed all of the content generated through the brand

guardrails (discussed above) AND a taste filter

Step #4: Have a real person with good judgment review, edit and

polish everything that comes out, so everything that gets

published is both valuable and relevant.

The fastest way to scale this is by building a library of proven

content structures. Formats like:

* Myth vs. Fact

* Buyer’s Guides

* Teardowns

* What to Expect

* Comparison Tables

* Checklists

These formats work because they speak to how people actually

evaluate solutions. They break down complexity. They frame

decisions. They let you communicate authoritatively without being

overly sales-y. Once built, each of these content types can be

reused and adapted across product lines, personas, or industries,

without reinventing the wheel each time.

This is how you scale velocity without sacrificing quality. AI

does the heavy lifting. Your team applies the judgment. Together,

that creates content that’s fast, on-brand, and grounded in real

user needs.

6. Start a Regular Portfolio Review

-----------------------------------

One of the benefits of portfolio-style management is that it

provides a structure for you to scale, maintain or deprecate

individual bets without undermining the initiative as a whole. As

you make this transition, one of the best things you can do is to

schedule time to review each set of bets you’re making

(efficiency, effectiveness, expansion) impartially – what’s

working? What’s not? Which processes does it make sense to invest

in automating? Which procedures should we leave alone?

It’s easy to automate something once, then forget about it – even

when that automation ceases to pay dividends or spawns unintended

consequences. The inevitable result of that approach is tech

debt, frustration and patchwork automations/workflows, in which a

new automation is created to fix the deficiencies of an existing

automation.

The superior approach long term is to treat every automation or

AI-driven workflow like a living asset, not a one-off project.

That means revisiting it regularly, assessing whether it’s still

delivering value and making an unbiased, clear-headed decision

whether it should be scaled, refined, or retired.

The reality is that - like any portfolio - some assets will

compound indefinitely. A reporting automation that saves five

hours a week for each of a dozen teams worth doubling down on.

Others will hit diminishing returns. Maybe the workflow only

works under certain conditions, or the maintenance overhead has

grown larger than the time it saves. At that point, it’s better

to deprecate it than keep propping it up.

What you’re building isn’t a collection of hacks. It’s a

portfolio. And like any portfolio, your job is to maximize

returns and minimize drag. That’s why regular (monthly/quarterly)

reviews matter - they keep you honest. They make sure you aren’t

using new workflows/tools as band-aids on top of broken ones.

All that leads to the real advantage: over time, you’ll grow a

portfolio of high-performing, margin-expanding automations,

workflows and tools while your competition winds up with

spaghetti workflows and staggering tech debt.

This week’s issue is brought to you by Optmyzr.

-----------------------------------------------

One of my favorite things about the tool is that - fundamentally

- Optmyzr isn’t about flashy features, it's about giving

marketers agency, clarity and leverage. I’ve demoed a ton of PPC

tools that try to take away control, and none of them work.

Optmyzr provides the best of both worlds: the automations (like

PPC Sidekick + Rule Engine) are powerful enough to run accounts,

while the platform itself is built around PPCers. The end result

is that you are still in control….but you have far more operating

leverage.

Here’s how Optmyzr maps to the portfolio framework we outlined:

* Efficiency (Low-Risk, Fast Payback): Automate repetitive tasks

with one-click optimizations, real-time alerts, rapid audits, and

quick-fire reporting — designed to free up time for strategy, not

just execution.

* Effectiveness (Decision-Support Upside): The AI Sidekick gives

immediate insights on where performance is shifting and why. You

get improved cause analysis and response speed without guesswork.

* Expansion (Upside in New Capabilities): Labs and AI-powered ad

text suggestions let you experiment rapidly with new creative

concepts and structures, with brand safety mechanisms built in.

Optmyzr handles tedious work so you can focus on narrative,

judgment and compounding advantage. If you want to see how

leveraging AI without surrendering control plays out in PPC,

Optmyzr offers a 14-day free trial (no credit card required).

-->Get Optmyzr (

link

)

Get Optmyzr ( link )

---------------------------------------------------

Your AI Strategy Should Read Like an Operating Plan

---------------------------------------------------

A good AI strategy isn’t a deck or a demo. It’s not a list of

tools, a set of GPT licenses or a pilot program. It’s an

operating plan that explains three things with ruthless clarity:

Where the leverage is. How you’re going after it. How you’ll know

if it’s working.

That plan should read like a portfolio, with efficiency bets that

quietly reduce friction, effectiveness bets that improve outcomes

and expansion bets that explore what might create new revenue

streams. Some will be successes, and some will be failures. The

discipline is in reallocating resources (capital, time, focus)

and attention regularly, so the portfolio as a whole keeps

getting stronger.

Everything else is theater.

So, as the noise ramps up this fall, skip the shiny objects.

Focus on workflows, leverage, distribution and the cadence of

review and reallocation. The winners won’t be the ones with the

most pilots or the flashiest announcements; they’ll be the ones

who quietly went about the boring work of using AI to build a

better, more efficient, more effective business.

And that’s the paradox. In a space obsessed with trends, sexiness

& hype, the real advantage belongs to the people/brands

disciplined enough to ignore it.

And when the inevitable question comes back — “What’s our AI

strategy?” — you’ll have an answer that isn’t just coherent, but

credible. One that moves the business forward, not just the

conversation.

Cheers,

Sam

Loving The Digital Download?

Share this Newsletter with a friend by visiting my public feed.

---------------------------------------------------------------

-->View the Newsletter Feed (

link

)

View the Newsletter Feed (

link )

Follow Me on my Socials

(

link

) (

link

) (

link

)

1700 South Road, Baltimore, MD 21209 | 410-367-2700

Unsubscribe (

link

) | Manage Preferences (

link

)

Want to read more from Sam Tomlinson?

Join Ads to AI to get full access to all 67 articles plus 500+ more from top AI and marketing thought leaders.

Join Ads to AI →