Choosing the right AI tools (for the right job)
It’s easy (when you know how) to build effective AI systems that genuinely work for your business – obviously, you need focus on the problem first – but when you’ve found the problem – how do you know what AI tools are right for the job?
Having spent a fair amount of time in the customer service space, I’ve witnessed firsthand how the same questions come up repeatedly. “I forgot my password,” “Where’s my delivery?” – these aren’t complex inquiries, but they eat up enormous amounts of time when handled manually. Each follows a predictable process: verify the customer, check the system, provide the standard solution. It’s no wonder every CCaaS vendor is peddling ‘AI’ as the meca-solution.
Today, I want to share what I’ve learned about building AI systems that actually work for business so you can appreciate the age old approach of Problem->Solution… (and counter the cuttently-popular AI tool -> problem). However, rather than getting lost in technical jargon, I’ll show you exactly which tools to use for which jobs, using real scenarios I’ve encountered.
The Problem: Building AI That Actually Solves Business Problems
Large Language Models like ChatGPT are impressive, but they have a critical limitation – they only know what they were trained on. Ask them about your company’s specific return policy or the status of order #12345, and they’ll give you generic responses or, worse, make something up entirely.
The real challenge isn’t just getting AI to be smart – it’s getting AI to be useful for your specific business needs. This means combining different AI tools in ways that match your actual workflows and requirements.
The AI Toolkit (a.k.a Team): Understanding Your Options
Think of building business AI like assembling a team. Each tool has a specific job, and the magic happens when you combine the right team members for your specific challenges.
Tool 1: RAG (The Research Assistant)
What it does: Searches through your knowledge base in real-time to find relevant information before answering questions.
Best for: When you have lots of documents and need accurate, up-to-date information.
Think of it as: Having an AI that can instantly flip through thousands of manuals, policies, and databases to find exactly what it needs.
Tool 2: CAG (The Walking Encyclopedia)
What it does: Preloads all relevant information directly into the AI’s memory beforehand, enabling instant responses.
Best for: When you have a manageable amount of stable information that doesn’t change often.
Think of it as: Someone who has memorized the entire company handbook and can give instant answers.
Tool 3: MCP Integration (The System Connector)
What it does: Connects AI to your existing business systems (CRM, payment processors, inventory) to access real-time data.
Best for: When AI needs current information from your live business systems.
Think of it as: Giving your AI direct access to check your actual systems rather than just documents.
Tool 4: Generative Media Creation (The Creative Engine)
What it does: Creates new content – text, images, videos – based on your requirements and data.
Best for: When you need to produce personalized or data-driven creative content at scale.
Think of it as: Having a creative team that never sleeps and can produce thousands of variations.
Tool 5: AI Agents (The Orchestrator)
What it does: Makes decisions about which tools to use when, and coordinates complex multi-step processes.
Best for: When tasks require planning, reasoning, and coordinating multiple tools.
Think of it as: A project manager that can think strategically and manage other AI tools.
Tool 1: RAG (The Research Assistant)
What it does: Searches through your knowledge base in real-time to find relevant information before answering questions.
Best for: When you have lots of documents and need accurate, up-to-date information.
Think of it as: Having an AI that can instantly flip through thousands of manuals, policies, and databases to find exactly what it needs.
Tool 2: CAG (The Walking Encyclopedia)
What it does: Preloads all relevant information directly into the AI’s memory beforehand, enabling instant responses.
Best for: When you have a manageable amount of stable information that doesn’t change often.
Think of it as: Someone who has memorised the entire company handbook and can give instant answers.
Tool 3: MCP Integration (The System Connector)
What it does: Connects AI to your existing business systems (CRM, payment processors, inventory) to access real-time data.
Best for: When AI needs current information from your live business systems.
Think of it as: Giving your AI direct access to check your actual systems rather than just documents.
Tool 4: Generative Media Creation (The Creative Engine)
What it does: Creates new content – text, images, videos – based on your requirements and data.
Best for: When you need to produce personalised or data-driven creative content at scale.
Think of it as: Having a creative team that never sleeps and can produce thousands of variations.
Tool 5: AI Agents (The Orchestrator)
What it does: Makes decisions about which tools to use when, and coordinates complex multi-step processes.
Best for: When tasks require planning, reasoning, and coordinating multiple tools.
Think of it as: A project manager that can think strategically and manage other AI tools.
Real-World Applications: Combining Tools for Different Jobs
Let me walk you through two specific scenarios – Customer Service and Digital Marketing – that illustrate when each approach shines.
Job 1: Simple Customer Service (Tool: CAG)
In my customer service days, I noticed that about 80% of inquiries fell into predictable categories:
“I forgot my password” – The process is always the same:
- Verify customer identity
- Send password reset link
- Provide backup authentication options if needed
“When are the bins collected?” (that’s residential trash collection for those outside of the UK) – Another standard workflow:
- Locate information in system
- Check resident address
- Provide bin collection information or escalate if there’s an issue
For these repetitive, process-driven interactions, CAG is the perfect tool. Here’s why:
- Lightning-fast responses: No time wasted searching through knowledge bases
- Consistent answers: Every customer gets the same accurate information
- Simple implementation: All your standard operating procedures can be preloaded
The right tool for the job: When your processes are standardised and your information is stable, CAG gives you the speed and consistency you need.
Job 2: Complex Customer Issues (Tools: AI Agents + MCP + RAG)
When customer service needs multiple tools: Here’s where it gets interesting. Sometimes a “simple” customer service question isn’t simple at all:
Customer: “I was charged twice for my order last week, but I only received one item. Can you help?”
The tool combination needed:
- AI Agent to coordinate the investigation process
- MCP integration to access your live business systems
- RAG to search through policies and procedures
This requires the system to:
- Access your CRM to verify the customer’s identity and order history
- Check the payment system to see the duplicate charges
- Query the inventory system to confirm what was shipped
- Review the refund policy to determine next steps
- Access shipping carrier APIs for delivery confirmation
Why you need multiple tools: No single tool can handle this complexity. You need an AI agent to orchestrate the process, MCP to connect to your systems, and RAG to search through policies – all working together.
Let’s switch to a different industry.
Job 3: Real-Time Performance Marketing (Tools: All of Them)
Now, let’s dive into the most sophisticated job: a digital marketing company leveraging first-party data for real-time performance optimisation. This requires the full toolkit working together.
The Challenge: Modern performance marketing isn’t just about running ads – it’s about creating a system that continuously learns, predicts, and optimises using your own customer data.
Tool Combination 1: Intent Prediction
- “This user just spent 3 minutes on our pricing page, viewed the enterprise tier twice, and downloaded our ROI calculator. What’s their likelihood to convert?”
Tools needed:
- MCP integration to access web analytics, CRM, and engagement data
- RAG to search historical conversion patterns
- AI agents to analyse and predict likelihood
Tool Combination 2: Dynamic Creative Generation
- “Create personalised video ads for high-intent enterprise prospects, incorporating our latest case study data.”
Tools needed:
- RAG to find relevant case studies and testimonials
- MCP to access real-time audience and performance data
- Generative media creation to produce video content
- AI agents to coordinate the entire process
Tool Combination 3: Automated Campaign Optimisation
- “Our demo signup campaign is underperforming in the 25-34 segment. Fix it.”
The complete process:
- MCP queries campaign APIs (Google Ads, Facebook, LinkedIn)
- RAG searches past successful campaigns for similar audiences
- AI agents analyse data and plan optimisation strategy
- Generative media creation produces new creative variants
- AI agents deploy and monitor new campaigns
- MCP provides real-time feedback for continuous learning
Why you need the full toolkit: Each tool handles what it does best – MCP connects to systems, RAG finds relevant information, AI agents make strategic decisions, and generative tools create content. Together, they create a marketing system that works autonomously.
This multi-step, reasoning-heavy process is exactly what Agentic RAG excels at. The AI agents can intelligently navigate between different data sources, compare information, and synthesise insights that a simple search-and-retrieve system couldn’t handle.
The MCP Advantage: What’s particularly exciting about Agentic RAG is how MCP integration allows it to connect directly to your existing business systems. Instead of just searching through static documents, it can query live databases, check real-time inventory, access customer transaction histories, and pull data from any system that exposes an API. This transforms AI from a document search tool into a true business intelligence agent.
Real-World Example: Your AI system detects that enterprise prospects who engage with case study content convert 3x better. It automatically generates personalised video case studies using your existing customer success stories, deploys them in targeted LinkedIn campaigns, and adjusts budget allocation based on real-time performance data – all while you sleep.
This represents the cutting edge of AI-powered marketing automation, where the system doesn’t just analyse and report – it predicts, creates, and acts autonomously by combining the right tools for each part of the job.
Now there’s a little more to it – but you don’t need to know about the ‘details’ (reach out to me instead) – and with advancements like googles new (at time or writing) ADK and the A2A protocol, things are getting easier to build despite all the technical change that’s happening.
Read googles ADK docs and learn more about A2A protocol (now part of the Linux Foundation.
Your Tool Selection Framework
Based on my experience implementing these systems, here’s how I recommend choosing the right tools for your specific job:
Simple, Repetitive Tasks:
Use: CAG
- You have a limited, stable knowledge base (standard operating procedures)
- Speed is critical (customer service, quick lookups)
- Your information rarely changes (company policies, product specs)
- You need consistent, repeatable responses
Perfect for: Customer service FAQs, internal policy queries, product information lookup
Complex Information Tasks:
Use: RAG (+ MCP if you need live data)
- You have a large, diverse knowledge base (thousands of documents)
- Your information changes frequently (news, real-time data)
- You need to search across multiple sources
- Accuracy is more important than speed
Perfect for: Research tasks, technical support, legal document analysis
Multi-Step, Strategic Tasks:
Use: AI Agents + Multiple Tools
- You need complex reasoning across multiple data sources
- Tasks require multi-step planning and decision-making
- You’re combining analysis with content generation
- Strategic thinking is required, not just information retrieval
- You need real-time access to business systems
- You’re doing performance marketing that requires predictive analytics
- Creative generation needs to be data-driven and personalised
Perfect for: Complex customer service issues, real-time performance marketing, automated campaign optimisation, personalised content generation
The Bottom Line
The choice between AI tools isn’t about which technology is “better” – it’s about matching the right tools to your specific business job.
From my customer service background, I learned that the best solutions are often the simplest ones that solve the immediate problem effectively. If you’re handling “forgot my password” requests, you don’t need a complex system with multiple tools – you need fast, reliable, consistent responses. That’s a CAG job.
But if you’re dealing with complex customer issues that require accessing multiple business systems, or building autonomous marketing systems that need to predict, create, and optimise continuously, you need multiple tools working together strategically.
The key insight: Start with the job, then choose the tools. Don’t get caught up in the technology – focus on what you’re trying to accomplish and build the simplest system that gets you there.
Getting Started
If you’re just beginning to explore AI tools for your business:
- Start with the job: Identify one specific business problem you want to solve
- Map your requirements: How much data? How often does it change? How complex is the task?
- Choose the simplest tool combination: Start with what meets your needs without over-engineering
- Test and iterate: Build small, learn fast, then expand
The AI landscape is evolving rapidly, but by understanding which tools do what jobs best, you’ll be well-equipped to make informed decisions that actually improve your business operations.
Remember: the goal isn’t to implement the most advanced technology – it’s to solve real business problems efficiently and effectively with the right combination of tools.
Here are some useful links to learn more:
Disclaimer – I have simplified the explanations and I have assumed your data (user guides, systems) are in an orderly state.
It’s always wise to speak to someone that can help you set up the stepping stones to delivering better solutions to your problems (Hi, did I mention that my name is Tatum).