--- title: "AI & Automation: Where to Start" description: "In our Spring 2026 Boxcar Marketing AI in Publishing & Marketing survey, we asked: \"If you could instantly get better at one thing using AI in your role, what would it be?\" A few themes emerged:..." url: https://www.boxcarmarketing.com/ai-automation-where-to-start/ date: 2026-06-17 modified: 2026-06-16 author: "Monique Sherrett" image: https://www.boxcarmarketing.com/wp-content/uploads/bluestonex-gLxNxONfRz0-unsplash-ai-automation.jpg categories: ["Harebrained Ideas", "Marketing Strategy & Tips"] type: post lang: en --- # AI & Automation: Where to Start In our Spring 2026 Boxcar Marketing (https://www.boxcarmarketing.com/ai-in-publishing-survey/), we asked: **“If you could instantly get better at one thing using AI in your role, what would it be?**“ A few themes emerged: saving time, improving reporting, writing more efficiently, choosing the right tools, and making content more accessible. To that end, here are some practical ways to start using AI to automate workflows. First off, let’s assume you have assessed that AI is appropriate for the task (a non-AI tool, program, or process may be better). Second, you have ensured that the risk of inputting confidential or sensitive information has been addressed, and training data is turned off in your tool of choice. ### 1. How can I use AI to automate more workflows? Start with the low-risk tasks you repeat often, not the tasks that are most annoying. Repetitive work is usually easier to automate, but it’s not always safer. Repeated tasks that involve sensitive judgment, exceptions, or rational nuance should not be your first automation attempt. Good candidates include: - **Turning meeting notes into task lists. **Lots of tools will help you do this. The key is ensuring those AI meeting notes are more than a forgotten asset. There are well-known commercial tools like Zoom, Microsoft Teams, Circleback, Notion, Granola, Otter, Fathom, as well as open-source, low-cost options for note-taking. But go one step further by integrating them into your project management or communication workflows: Asana, Monday.com, Salesforce, or Hubspot. Additionally, ChatGPT and Claude both let you add Connectors (aka integrations) to your meeting note taker. Once linked, these AI tools can then pull context from your note taker into chats, or Projects. - **Drafting first-pass campaign summaries.** This is a grey-area case where you’d want to make sure you have training data turned off AND approval to screenshot and input campaign data into an AI tool. What does approval look like? Ideally, your organization has a clear AI policy that outlines what categories or data can be used, what needs approval first from an established point-person, and what steps you’re taking regarding compliance, consent, data retention, or client/customer confidentiality. If it’s a go, instead of looking at your campaign reports directly in the platform—Google Ads, Meta Ads, Mailchimp—you can screenshot those reports and ask Claude or ChatGPT questions to help you get started drafting your summary of the campaign performance. “Based on these results, what are three specific insights I should highlight in an executive summary? Are there any anomalies that require investigation? What can you tell me about this campaign, and what CAN’T you tell me about it? What’s missing from this data that you typically see in a healthy campaign?” Remember that the AI can only see what’s in the screenshot so the insights are limited, but as a first-pass these types of questions can surface interesting directions to explore. The insights will typically sound credible so don’t rely on this, use it as a starting point. - **Creating social post variations from approved copy**. In Claude or ChatGPT you can create a Project that has several files that outline your brand voice, top-performing posts on different platforms, and any type of template or style guide that you follow. Then in that Project, you can ask the AI Tool to convert a core idea or long-form post into platform-specific lengths (e.g., Facebook post, Instagram reel, LinkedIn stories). Hootsuite and Buffer have integrations with ChatGPT and Claude that can also help you manage content and draft posts. A simple first step is to document the workflow before automating it. Write down: trigger, input, steps, output, and who needs to review it. Once the process is clear, AI can help draft templates, summarize information, or move content between tools. **1-minute marketing tip:** Before automating a task, write the recipe. AI works better when your process is clear. My favourite approach is to ask AI to help me understand how to do what I want vs. just giving it instructions. Get the tool to teach you how to use it best. ### 2. What’s a practical way to use AI for reporting and data analysis? MCPs (or Model Context Protocol) allow AI agents (like Claude and ChatGPT) to be configured to interact securely with analytics and advertising APIs, depending on permissions and controls. Instead of exporting and uploading CSV files or taking screenshots as discussed above, you can ask questions about your data within the chat. Interpreting analytics and ad reports can be a struggle for marketers who have not been trained in data analysis. So instead of just using the AI to make charts, ask the AI to help you ask better questions of your data. For example, instead of asking, “How did this campaign do?” try: - Which channels drove the strongest engagement? - Which subject lines or ad messages performed best? - Did publicity, email, social, and paid campaigns reinforce each other? - What should we repeat, stop, or test next time? AI can summarize GA4 data, email performance, ad results, social metrics, and sales notes, but the quality of the answer depends on clean inputs and good context for goals, audience, timeframe, and known anomalies. Always explain the campaign goal, the audience for the report you are creating, and any important context, including the reporting timeframe. Before leaning on AI, the big question you should start with is, “what action will we take if performance is high vs. low?” Understanding the data should be tied to some action that has an impact on the business. **1-minute marketing tip:** Don’t ask AI to “analyze the report.” Ask it to answer a specific marketing question. “How can I assess if this campaign actually drove new value, or if we paid for conversions that would have happened organically?” Or, “Do we have enough information to understand how the various touchpoints in a multi-channel campaign are assisting each other along the funnel?” And my ultimate question is always to ask about what we cannot infer from the provided data. Most of the time, the AI cannot infer causality from ordinary reports, but if you can get suggestions on the types of reports to run, or the baseline modeling needed, or the attribution design required for future campaigns, then those tips are valuable. ### 3. Can AI help with campaign reports? Yes. AI is especially useful for turning campaign activity into a clear narrative. A useful campaign report structure is: - What we were trying to achieve - What we did - What happened - What we learned - What we recommend next For publishers, this can be helpful when combining multiple campaign signals: email clicks, media coverage, event activity, influencer posts, retailer promotions, website traffic, and sales notes. AI can help draft the report, but a person should still check the interpretation, especially when sales data is incomplete or timing is hard to attribute. Anecdotally, I’ve had experiences where the data has been misinterpreted or conflated. I’ve had experiences uploading CSV files and the AI does not use the whole file, only a sample. Basically I do not find AI useful for analytics (yet), but I’ve only started experimenting with MCPs, and I have not systematically tested my examples across tools or with enough different file sizes. I do find AI helpful for creating a templated structure that I can use in reports. In a Project, I upload examples of reports that I find useful and actionable. Then I ask the AI to identify the shared structure, the Pros and Cons of each approach, and to recommend a template that suits my particular context, audience, and desired output. **1-minute marketing tip:** Use AI to draft the story of the campaign, then use your judgement to confirm what the data really supports. ### 4. How can AI help with spreadsheets? AI can help with spreadsheet efficiency by writing formulas, cleaning messy data, categorizing rows, creating summaries, or explaining what a formula is doing. Useful prompts include: - Write a formula to extract the publisher identifier in the ISBN from this cell. - Explain this formula in plain language. - Suggest a pivot table structure for this sales report. - Find inconsistencies in these title names. - Detect outliers and anomalies in numeric columns. Another use case is converting Excel to Google Sheets. Sometimes Google Sheets can drop or mess up Excel-created functionality on import. I had an Excel report with a ranking formula that worked across multiple columns of data. When I tried to convert it to Google Sheets, it would not work and I couldn’t figure out what was wrong. So I had ChatGPT create an Extensions > Apps Script that I could run in Google Sheets, which I used instead of File > Import. The Apps script created the tabs, dropdowns, ranking formulas, escalation logic, and recommendations. Again, be careful with private or sensitive data. Remove customer information, personal emails, and confidential sales details before using AI tools. Some formulas, like extracting an ISBN, are low-risk and do not involve sensitive data. If you’re writing formulas for analysis of confidential or financial data then the input is sensitive and erroneous calculations can have serious impact. **1-minute marketing tip:** AI is great at spreadsheet cleanup, but don’t upload sensitive data unless you understand the tool’s privacy settings. ### 5. How can AI help condense sales handles or marketing copy into 1–2 sentences? AI is good at creating options, but you still need to provide the positioning. This is where experienced marketers excel because they already understand the positioning and have access to accurate comparable titles, audience insights, and sales angles. Be sure to include that information in your prompt. Less experienced marketers may produce confident but poorly grounded copy if they don’t do the research in advance. A strong prompt includes: - The book’s audience - The comparable titles or authors - The emotional hook - The format or genre - Any sales angle, award, review, or timely connection For example, describe the above in your prompt, then request the following: “Condense this book description into a 2-sentence sales handle for booksellers. Keep the tone warm, specific, and benefit-focused. Avoid hype.” Then ask for 5 versions: one for booksellers, one for media, one for librarians, one for social, and one for internal sales reps. This helps you see which angle is strongest. I also ask for a confidence score and an explanation of why the AI thinks its strongest pick is best. Understanding the reasoning helps you understand the tool and what it’s doing. It’s also helpful for refining your marketing tactics. **1-minute marketing tip:** Don’t ask AI for “better copy.” Ask for copy for a specific reader, channel, and job. And, provide as much context as possible. (https://docs.google.com/document/d/1_nn9FAhGjGyXbDVjKii3uYvIxjJz2bBa2DLG6uw_9Ro/edit?usp=sharing) ### 6. How can I keep my brand voice while using AI for writing? Start by giving AI examples of your best writing. Ask it to identify the voice: sentence length, tone, vocabulary, rhythm, humour, formality, and words you tend to avoid. Then create a short voice guide you can reuse. For example: Write in a professional, friendly tone. Be practical, clear, and lightly conversational. Avoid jargon, exaggerated claims, and overly polished corporate language. This is especially helpful if AI supports your writing process by improving spelling, structure, or clarity. You can ask AI to clean up your draft while preserving your ideas and phrasing. Try prompts like: - Edit this for clarity and spelling, but keep my voice. - Make this more coherent without making it sound generic. - Keep the sentence structure close to mine, but improve flow. - Flag any edits that change the meaning. Regular AI users also know to move this type of conversation from the basic chat into a Project or a Skill, where you have a brand voice document, and examples of your best work, a list of do’s and don’ts and a style guide, vocabulary guide, words or jargon to avoid, etc. Skills and Projects make these processes repeatable. Start in the basic chat and ask it to help you develop a brand voice document and identify the assets need for a Skill or Project, then test-run a few times to correct anything, and go from there. And remember to monitor and update the skill or project as needed. As your writing style and voice evolve, you do not want to preserve outdated or overly narrow brand voice patterns. **1-minute marketing tip:** Your voice matters. Use AI as an editor, not a ghostwriter. ### 7. What tools should I use for which AI jobs? It’s easy to rely on one tool for everything, but different tools are better at different jobs. A practical way to think about it: - **ChatGPT, Claude, or Gemini:** brainstorming, drafting, editing, summarizing, planning, and analysis - **Built-in AI tools in Google Workspace or Microsoft 365:** email, docs, meeting notes, and internal productivity - **Canva or Adobe tools:** design support, image resizing, layout variations, and visual content - **Zapier, Make, or native integrations:** workflow automation between platforms - **GA4, Looker Studio, spreadsheet tools:** reporting, dashboards, and structured analysis - **Accessibility tools and human review:** alt text, transcripts, captions, and quality control The best tool depends on the job, task type, risk level, budget, user skill, accessibility, data sensitivity, integration or auditability. For most publishing teams, the first priority is not finding the fanciest tool. It’s matching the tool to a real workflow. And the major AI brands mentioned above are not the only option. Look for vendor fit based on your governance, accessibility requirements, cost, and data-residency needs. **1-minute marketing tip:** Choose the AI tool after you define the task, not before. ### 8. Can AI help with alt text for complex images, like sheet music? AI can help create a first draft of alt text, but complex images need careful human review by someone with accessibility or subject-matter expertise. Sheet music, charts, diagrams, maps, and illustrated spreads often contain information that AI may miss or misinterpret. For complex images, start by asking: “What does the reader need to understand from this image in this context?” For sheet music, useful alt text may need to describe the type of notation, instrument, mood, structure, or relevant musical markings. But if the notation itself is the content, alt text alone may not be enough. You may need a longer description, caption, transcript, or accessible supplement. **1-minute marketing tip:** For complex images, describe the purpose of the image, not just what it looks like. ### 9. What’s the best first automation for a busy marketing or publicity team? Start with an automation that reduces admin work but keeps humans in control. “Human in the loop” shouldn’t be symbolic. Define what needs review, who reviews it, what authority/expertise they need to have, and what failure points should be checked. Good first projects include: - Turn media coverage links into a weekly roundup draft - Convert author event details into a promo checklist - Summarize sales conference notes by title - Generate follow-up tasks from meeting notes Avoid starting with anything high-risk, such as automatically sending emails, publishing posts, changing live website content, or making decisions without review. **1-minute marketing tip:** Automate drafts and checklists before you automate decisions. ### 10. What should we be cautious about? AI can save time, but it can also make mistakes confidently. Treat AI outputs as drafts, not final answers. Watch for: - Generic copy - Incorrect facts - Misread data - Lost brand voice - Privacy risks - Over-automation of work that needs judgement - Accessibility descriptions that miss the real meaning of an image A good rule of caution: the more public, sensitive, or strategic the output is, the more human review it needs. **1-minute marketing tip:** AI can speed up many processes. It should not replace human judgement, especially at the beginning (what are we doing and why, is this actually an AI task or better done by a human or other tool) and at the end (validate the output, handle the corrections, train the model on required future outputs). ## Final Takeaway The strongest AI opportunities for publishing teams are not about replacing creative or strategic work. They are about reducing tasks, improving clarity, speeding up reporting, and giving teams more time to focus on judgement, relationships, and better marketing decisions. For seasoned marketers, AI can free up capacity, but it can also intensify the workload, result in role changes, and unevenly impact other team members. Start small. Pick one low-risk repeated task. Define the workflow. Test the output. Update and review. Then improve and build from there.