AI Crossroads: One way to choosing the right AI Stack for your Marketing Team
There is enormous interest in bringing AI into marketing operations—and understandably so. In the past years, everyone I talk to is facing similar realities: marketing teams are being tasked to produce more campaigns, more channel variations, more reporting, making sure customer experiences are more unique and personalized, often at flat budgets and without access to additional resources. AI seems to be the solution as it promises faster execution, lower costs and the ability to scale work that previously required substantial time or headcount.
Unfortunately, many AI conversations seem to begin with the wrong question:
What work can we fully offload, which processes replace or automate using AI?
That approach may produce an attractive short-term business case, but it often misses a larger opportunity.
In my experience, marketing is rarely slowed down because someone cannot write a first draft quickly enough. It is usually slowed down by incomplete briefs, scattered information, uncertain ownership, duplicated work, approval bottlenecks, disconnected systems and decisions waiting in someone’s inbox.
Replacing a workflow or even worse, team members, does not fix those problems.
A better starting question is:
What outcome are we trying to improve, and what prevents the team from achieving it today?
That question changes both the technology decision and the organization’s approach to AI. The goal should not be to make every employee produce twice as much content. It should be to help the team make better decisions, eliminate low-value administrative work and move strong ideas through the organization more effectively.
Efficiency is not the same as output
AI can create an impressive amount of material. It can produce headlines, campaign concepts, email variations, customer segments, social posts, research summaries and images in seconds. But increasing output is only useful if the organization can review, approve, activate and learn from that output.
Otherwise, AI simply moves the bottleneck. A creative team may generate fifty campaign variations instead of five, but legal still needs to review them. Marketing may create more leads, but sales still needs to qualify and follow up with them. Teams may receive faster reports, but someone still needs to decide what action to take.
This distinction matters:
Productivity means completing an existing task faster.
Efficiency means removing unnecessary effort from the wider process.
Effectiveness means improving the final business outcome.
AI investments should ultimately be judged by the third definition.
A faster campaign process is valuable. A faster process that produces better-performing campaigns, clearer decisions and less rework is considerably more valuable.
Begin with the workflow—not the tool
Almost every marketing team I know and work with is already experimenting with AI.
One person uses ChatGPT for campaign concepts. Someone else uses Gemini to summarize research. An agency presents AI-generated visuals. Sales has turned on an AI feature inside the CRM, while operations is quietly building automations with models like Zapier.
Each experiment may save time individually. Collectively, however, they can create a new layer of disconnected tools, subscriptions and information. For a marketing team of ten or more—particularly one working with agencies, sales, product, operations and finance—the selection criteria need to extend beyond the quality of an AI-generated answer.
Before evaluating platforms, map the workflow:
Where does the work begin?
What information is required?
Where is that information stored?
Who makes the decisions?
Where do delays and errors occur?
Which steps require human judgment?
What would a meaningfully better outcome look like?
Only then should the team decide where AI belongs. Sometimes the answer will be a general AI workspace. Sometimes it will be an automation inside the CRM. Sometimes it will be a better intake form, a clearer approval process or simply one agreed source of truth. Not every operational problem needs an AI solution.
The four layers of a useful AI stack
A model such as GPT, Gemini or Claude may provide the reasoning engine, but a model alone does not create an efficient marketing operation.
A practical team solution typically requires four layers:
A general AI workspace for research, analysis, writing and problem-solving.
A system-of-record layer connected to real customer, product and sales data.
A workflow layer that moves information between applications.
A governance layer covering permissions, brand standards, approvals and accountability.
The objective is not to purchase a separate AI application for every marketing activity. It is to create a small, understandable stack in which each platform has a clear role.
Choose one primary AI workspace
Most organizations do not need ChatGPT, Microsoft Copilot, Gemini and Claude licensed for every employee.
They need one approved primary environment that matches the way the organization already works.
Bring AI to your customer data—do not create a second CRM
Some of the most valuable marketing uses of AI have little to do with generating content.
AI can help with:
Lead enrichment and routing
Customer segmentation
Sales and service handoffs
Campaign-performance analysis
Identifying missing customer information
Preparing personalized follow-up
Detecting customers who may need attention
These applications depend on reliable customer data. For that reason, the AI already connected to the organization’s CRM will often deliver more operational value than a separate specialist tool.
Introducing some AI tools I came across so far
ChatGPT Business: strong across a mixed technology stack
ChatGPT Business is a capable general-purpose option for strategy, research, data analysis, content development and cross-functional problem-solving.
Its strength is flexibility. It is useful when the team works across different platforms rather than operating entirely inside Microsoft or Google. Connected company knowledge can also help employees find and synthesize information from approved internal sources.
At around US$20 per user per month, a ten-person team starts at approximately $200 per month. It includes a shared workspace, administrative controls and access to advanced OpenAI tools and models. OpenAI pricing
It is particularly useful for:
Strategic and campaign planning
Market and competitor research
Working across documents and data
Creating reusable internal assistants
Teams with a mixed technology environment
The risk is treating it as a new location for unofficial customer, product or campaign records. The CRM, ERP and project-management system should remain the authoritative sources.
Microsoft 365 Copilot: the natural fit for Microsoft organizations
If work already happens in Outlook, Teams, SharePoint, Word, Excel and PowerPoint, Microsoft 365 Copilot may offer the most direct path to adoption.
Its primary advantage is proximity to the organization’s daily work. It can summarize meetings, retrieve internal information, prepare documents and analyze spreadsheets without requiring employees to continually transfer material into another application.
Microsoft also provides more than 100 connectors for external platforms, including Salesforce, ServiceNow, Box and Confluence. Microsoft’s connector overview
The Copilot add-on is listed at approximately US$30 per user per month, in addition to a qualifying Microsoft 365 subscription. Microsoft 365 Copilot pricing
That cost may be justified for employees who spend much of their day coordinating work through Microsoft applications. It is harder to justify when licenses are distributed widely without defined use cases.
Google Workspace with Gemini: an attractive integrated value
For organizations already working in Gmail, Drive, Docs, Sheets and Meet, Gemini offers a cost-efficient route to integrated AI.
Google Workspace Business Standard is listed at US$14 per user per month on an annual commitment and includes Gemini functionality in core Workspace applications. Google Workspace pricing
It is a strong option for:
Collaborative research and drafting
Email and meeting summaries
Working with information stored in Drive
Agency and client collaboration
Organizations looking to limit incremental AI costs
As with every knowledge tool, its effectiveness depends on the quality and organization of the underlying information. AI can search a disorganized Drive, but it cannot automatically establish which of six similarly named campaign documents is correct.
HubSpot Breeze: a practical mid-market option
For organizations already using HubSpot, Breeze adds AI to existing marketing, sales and service processes without establishing another customer-data environment.
Breeze can research companies, enrich records, support outreach and automate structured tasks using HubSpot data. Its assistant is included with HubSpot subscriptions, while more advanced agents and actions may consume HubSpot Credits. HubSpot Breeze overview
HubSpot currently lists completed-outcome pricing of $0.50 per resolved Customer Agent conversation and $1 per lead prepared by its Prospecting Agent. HubSpot AI pricing
This type of outcome-based pricing is promising, but it still needs to be measured against lead quality, resolution quality and the cost of human review.
Salesforce Agentforce: powerful within an established Salesforce environment
Agentforce is most compelling when Salesforce is already the centre of the organization’s customer data, sales processes and marketing infrastructure.
It can help coordinate campaigns, personalize customer interactions, update records and automate work across marketing, sales and service. Salesforce offers consumption-based, conversation-based and per-user pricing options. Salesforce Agentforce pricing
For a large Salesforce customer, this can provide a deeply integrated foundation. For an organization with fragmented customer data or inconsistent processes, however, Agentforce will not be a quick installation.
The quality of any AI output will remain dependent on the quality of the underlying data and operating model.
The workflow layer is where AI becomes operational
A capable model can recommend what should happen. A workflow platform helps make it happen consistently.
Consider a common campaign-request process:
A department submits a campaign request.
AI checks whether the brief contains the required information.
The request is classified and routed to the correct team or agency.
A project, timeline and approval path are created.
Relevant customer, product and brand information is retrieved.
The team develops the campaign.
A person reviews and approves the work.
Approved assets are activated.
Results return to the reporting environment.
This is more valuable than simply using AI to write the campaign brief faster.
Zapier: accessible automation across a mixed stack
Zapier is one of the easiest automation platforms for marketing and operations teams to adopt. It connects more than 9,000 applications and can move information between AI models, forms, CRM platforms, project-management tools and communication systems. Zapier platform overview
It is well suited to:
Lead capture, enrichment and routing
Campaign and approval notifications
Transferring information between agencies and internal teams
Recurring reporting
Testing automation ideas without a lengthy development project
The main concern is cost and governance at scale. Usage-based task charges can grow, while undocumented automations can become difficult to maintain.
Make: greater visibility for more complex automation
Make offers a visual environment for building multi-step workflows and agents across more than 3,000 applications. It can provide more control over branching, data transformation and error handling for teams comfortable designing operational processes. Make plans and capabilities
Regardless of platform, each production workflow should have:
A named owner
Written documentation
Defined input and output
Testing before significant changes
Error alerts
A manual fallback
Regular cost and performance reviews
An automation that nobody owns eventually becomes a liability.
Connecting AI to ERP systems requires greater care
There is understandable interest in connecting marketing AI to inventory, pricing, order, finance and product information held in ERP systems.
This can create valuable capabilities. A marketing team could, for example, prevent the promotion of unavailable products, tailor campaigns to regional inventory or compare marketing spend with actual margin.
It also introduces much greater risk.
Platforms such as SAP, Oracle, Microsoft Dynamics and legacy ERP systems contain sensitive operational data and critical business rules. AI should access this information through approved, controlled connections—not through improvised integrations with broad permissions.
Organizations working with ERP or legacy systems should consider established integration environments such as MuleSoft, Boomi or Microsoft Power Platform. The marketing team should work with IT, operations, finance and security to define exactly what information the AI can read and which actions it can take.
Access to data and authority to change data are very different permissions.
Specialist tools should solve specialist problems
Additional AI platforms become worthwhile when the scale or complexity of the work justifies them.
Adobe GenStudio for high-volume creative operations
Adobe GenStudio for Performance Marketing is designed to help organizations create, adapt, approve and measure large numbers of branded campaign assets.
It integrates with Adobe Experience Manager, Express, Firefly, Journey Optimizer, Marketo and Campaign, as well as platforms including Meta, LinkedIn and TikTok. Adobe GenStudio integrations
For a global organization producing hundreds of asset variations across markets and channels, that can address a real operational need.
For a smaller team producing a manageable number of campaigns, it may be more infrastructure than the workflow requires.
WRITER for brand and compliance governance
WRITER is designed for organizations that need controlled AI content workflows, consistent terminology and formal brand or legal review. Its agents can connect with enterprise systems and include human approval points. WRITER’s marketing platform overview
It becomes relevant when many departments or agencies are producing content, particularly in multi-brand, multilingual or regulated environments.
It may be unnecessary when a single team has relatively straightforward review requirements.
Asana AI Studio for intake and work coordination
When Asana already manages marketing work, AI Studio can improve the operational layer by checking briefs, categorizing requests, routing tasks, summarizing projects and escalating exceptions.
This is a good example of enabling a team rather than replacing it. AI handles the administrative coordination while people retain responsibility for priorities, ideas, relationships and decisions.
The replacement trap
There will be workflows where AI reduces the amount of labour required. Pretending otherwise would be unrealistic. But beginning an AI initiative with a headcount-reduction target creates several problems:
First, employees have little incentive to identify meaningful automation opportunities if success appears likely to eliminate their positions.
Second, replacement-led programs tend to focus on visible tasks rather than underlying processes. Writing a product description is visible. Finding the correct product information, confirming inventory, coordinating approvals and updating every sales channel is the actual workflow.
Third, the work AI performs still needs ownership. Someone must define the objective, provide context, evaluate the result, handle exceptions and remain accountable.
Finally, reducing capacity too early can eliminate the people who understand why the process works the way it does. That knowledge is often essential to redesigning it successfully.
A more productive objective is:
Remove repetitive work, strengthen decisions and increase the capacity of good people to create better outcomes.
The eventual organizational design may change. But it should change because the work has been thoughtfully redesigned—not because a technology demonstration made replacement appear easy.
Try this 3 step practical approach
Step 1: Diagnose before buying
Identify three recurring workflows that create frustration across teams or departments.
Useful candidates include:
Campaign intake and briefing
Customer or market research
Lead qualification and routing
Agency coordination
Campaign reporting
Meeting follow-up
Content review and approval
Record the current time, number of handoffs, rework, delays and common errors. Define the desired outcome before discussing products.
Step 2: Design controlled pilots
Select the smallest combination of existing tools that can improve the workflow.
Define:
What information the AI can access
Which steps are automated
Where human judgment remains essential
Who approves the final result
What the system is prohibited from doing
How errors are identified and corrected
Who owns the workflow after launch
Employees doing the work should participate in designing the pilot. They usually know where the real friction—and the dangerous exceptions—exist.
Step 3: Measure, learn and standardize
Compare the pilot with the original process using measures such as:
Turnaround time
Hours of repetitive work removed
Rework and error rates
Cost per completed workflow
Employee adoption
Lead, campaign or customer outcomes
Quality of cross-department handoffs
Expand what demonstrably improves the outcome. Redesign what creates new problems. Retire the experiments that produce activity without value.
Where independent guidance helps
The AI market is evolving too quickly for most organizations to evaluate every platform, integration and pricing model internally.
Vendors are naturally focused on what their products can do. The organization still needs someone to determine what it should do.
Depending on where your organization is in the process, that may include:
Auditing current marketing workflows and identifying meaningful AI opportunities
Separating process problems from technology problems
Prioritizing opportunities by impact, effort and risk
Evaluating tools independently of vendor sales claims
Designing a practical AI and automation stack around existing systems
Developing pilot workflows with clear human approval points
Coordinating marketing, sales, agencies, IT and other stakeholders
Establishing practical governance and ownership
Supporting training, adoption and change management
Measuring whether the investment improves an actual business outcome
Providing interim or fractional leadership while the new approach is implemented
The objective is not to prescribe the largest platform or the most ambitious automation.
It is to help the organization make thoughtful choices, prove value early and build an approach its people can understand and sustain.
The final test: does the team function better?
AI adoption should not be judged by the number of licenses purchased, prompts submitted or pieces of content generated. It should be judged by whether the organization works better.
Before approving another AI platform, ask:
What outcome are we trying to improve?
What currently prevents the team from achieving it?
Does this tool fix the constraint or merely accelerate one task?
Does it work with systems we already trust?
Does it enable employees to contribute more value?
Can we measure the improvement?
Who remains responsible when the system gets something wrong?
The winning AI stack will not be the one with the most tools or the boldest promise of automation. It will be the one that removes friction, strengthens the people doing the work and helps good decisions move through the organization more effectively.
I am always interested to hear how people integrate AI in their workspace. Hit me up if you want to exchange best practices, need advice or looking for ways to improve to the ways your business shows up in the world.
Comments