Why Every SMB Needs a Data Pipeline Strategy
Previous Post Your CRM has customer data. Your accounting software has financial data. Your project management tool has workload data. Your email has communication data. Your spreadsheets have… everything else. None of them talk to each other. This is the data silo problem, and it’s one of the biggest invisible drags on business performance. The solution? A data pipeline strategy. What Is a Data Pipeline? A data pipeline is an automated system that moves data from one place to another, collecting it from sources, transforming it as needed, and delivering it to a destination where it can be used. Think of it as plumbing for your data. Instead of someone manually exporting a CSV from your CRM, reformatting it in Excel, and uploading it to your reporting tool every week, a pipeline does this automatically, reliably, and without errors. Why SMBs Need This More Than Enterprises Large enterprises have data engineering teams. They build and maintain complex data infrastructure as a core function. Small and mid-sized businesses typically don’t have this luxury, but they have the same problem. Data lives in multiple tools that don’t sync. Decisions are made on incomplete or outdated information. Team members spend hours each week manually moving data between systems. The irony is that SMBs often have simpler data ecosystems, which means building pipelines is faster, cheaper, and delivers proportionally bigger impact. The Cost of No Strategy Without a data pipeline strategy, here’s what happens. Duplicate data entry The same customer information is entered into three different systems manually. Each version drifts slightly out of sync over time. Stale reports Your monthly report is based on data someone pulled last Tuesday. By the time decisions are made, the numbers are already outdated. Missed signals A customer who cancelled their subscription last week is still receiving upsell emails because the CRM wasn’t updated. An overdue invoice goes unnoticed because finance checks a different system. Wasted time Someone on your team, probably one of your best people, spends several hours each week as a “human API,” manually transferring data between tools. Building Your First Data Pipeline You don’t need a massive infrastructure project. Start with these steps. Step 1: Map your data sourcesList every tool your team uses and what data lives in each one. Identify which data needs to flow between systems. Step 2: Identify the highest-value connectionWhere does manual data transfer cause the most pain? That’s your first pipeline. For most SMBs, it’s CRM-to-accounting, form-to-database, or multi-source-to-report. Step 3: Choose your tools For Microsoft-heavy environments, Power Automate and Azure Logic Apps cover most use cases. For custom or complex needs, Python scripts running on Azure Functions provide maximum flexibility. Step 4: Build, test, monitor Build the pipeline, test it with real data, and set up monitoring so you know immediately if something breaks. Automated pipelines should be transparent; you should always know what data was moved, when, and whether it succeeded. Step 5: Expand graduallyOnce the first pipeline is running, expand to the next highest-value connection. Over time, you build a network of automated data flows that keeps your entire business in sync. The Result A well-designed data pipeline strategy gives you a single source of truth across your tools, real-time (or near-real-time) data in your reports, zero manual data transfer, fewer errors and inconsistencies, and more time for your team to focus on analysis and action instead of data wrangling. Start Small, Scale Smart You don’t need to automate everything at once. The beauty of data pipelines is that they’re modular — each one solves a specific problem, and together they create a connected, efficient operation. Ready to connect your data? Let’s map your pipeline strategy together.
Automating SharePoint Workflows with Power Automate
Previous Post If your company uses SharePoint, you’re sitting on a massive automation opportunity, and Power Automate is the key to unlocking it. SharePoint is where most Microsoft 365 organizations store documents, manage lists, run team sites, and track information. Power Automate connects directly to SharePoint with deep, native integration. Together, they can automate some of the most common and time-consuming workflows in your business. The Power Couple Power Automate has over 30 SharePoint-specific triggers and actions. You can trigger a flow when a file is created or modified, when a list item changes, when a folder is updated, or based on custom conditions. Then you can automatically move files, send notifications, update metadata, create approvals, or push data to other systems. Here are the workflows we build most often: Document Approval Workflow A new document is uploaded to a SharePoint library. Power Automate detects it, sends an approval request to the designated reviewer (via Teams or email), and waits for a response. Once approved, the document’s status column is updated, the author is notified, and the file is optionally moved to a “Published” folder. This replaces the classic “I emailed it to you, did you see it?” cycle. Automated Metadata Tagging When documents are uploaded, a flow can automatically tag them with metadata based on the folder path, file name, or even the content. This keeps your document libraries organized and searchable without relying on users to fill in metadata manually, which they rarely do. New Item Notifications A SharePoint list tracks project tasks, IT requests, inventory items, and client orders. When a new item is added or an existing item’s status changes, Power Automate sends targeted notifications to the right people via Teams, email, or SMS. No more checking lists manually for updates. Form-to-SharePoint Pipelines Microsoft Forms or custom forms collect data (job applications, feedback surveys, order requests). A Power Automate flow captures each submission, creates a new item in a SharePoint list, attaches any uploaded files, and notifies the responsible team. The data is immediately organized and actionable. Scheduled Reporting A flow runs every Monday morning, queries a SharePoint list for items matching certain criteria (overdue tasks, pending approvals, items added this week), compiles the data into a formatted email or Teams message, and sends it to the manager. Consistent, reliable reporting with zero effort. File Lifecycle Management Old files in SharePoint? A flow can scan libraries on a schedule, identify documents that haven’t been modified in 6+ months, notify the owner to review or archive, and automatically move neglected files to an archive library. This keeps your SharePoint clean and your storage costs down. Tips for Better SharePoint Automation Keep your SharePoint lists and libraries well-structured — clean columns and consistent naming make automation much easier. Use SharePoint content types to standardize document metadata across libraries. Test flows with a dedicated test library before deploying to production. Use flow error handling and configure notifications for failures. Getting Started If you’re already using SharePoint and Power Automate is included in your Microsoft 365 plan, you have everything you need to start. The ROI on these automations is typically immediate, most can be built and deployed in a single day. Need help building SharePoint automations that actually work? We do this every day.
How to Build a Custom AI Assistant for Your Team
Previous Post Your team has questions. Dozens of them, every day. “What’s our refund policy for enterprise clients?” “Where’s the latest version of the brand guidelines?” “What was the revenue for Q3?” “How do I submit an expense report?” Right now, the answer to these questions involves digging through SharePoint, searching old emails, pinging colleagues on Slack, or waiting for someone who “just knows” to be available. A custom AI assistant changes this entirely. What Is a Custom AI Assistant? It’s an AI-powered interface accessible via Teams, Slack, a web app, or any channel, where your team can ask questions in natural language. Unlike a basic search tool, it understands context, synthesizes information from multiple sources, and gives direct answers. The keyword is “custom.” This isn’t a generic chatbot. It’s trained on your company’s specific knowledge: your documents, policies, processes, data, and terminology. It knows your business. What Can It Do? Answer knowledge questions “What are the payment terms in our standard vendor contract?” The assistant retrieves the relevant contract template, reads it, and responds with the specific terms with a link to the source document. Summarize documents “Give me a summary of last month’s board report.” Instead of reading a 30-page PDF, your team gets a concise summary in seconds. Surface data “How many support tickets did we close last week?” If connected to your ticketing system, the assistant can pull real-time metrics. Guide processes “How do I request a new laptop?” The assistant walks the user through your IT procurement process step by step, based on your internal documentation. Draft content “Write a first draft of the client update email based on this week’s project notes.” The assistant generates a draft using your preferred tone and format. How It’s Built Building a production-ready AI assistant involves several components: Knowledge base Your company documents, policies, procedures, guides, FAQs, and reports are processed, chunked, and stored in a vector database. This is the foundation of the assistant’s knowledge. Retrieval system (RAG) When a user asks a question, the system searches the vector database for the most relevant document chunks, then passes them to the AI model along with the question. This ensures the assistant answers based on your actual data, not generic internet knowledge. Language model The AI model (hosted on Azure OpenAI or similar) generates the response based on the retrieved context. It’s configured with your business rules: what it should and shouldn’t answer, how formal to be, when to escalate to a human. Interface The assistant is deployed wherever your team works, Microsoft Teams, Slack, a web portal, or a custom app. Security and access control Different teams see different data. The assistant respects your existing permission structure so that sales can’t accidentally access HR documents and vice versa. What It Takes TimelineA basic internal assistant can be live in 2–3 weeks. More complex setups with multiple data sources, custom integrations, and role-based access typically take 4–6 weeks. Data requirementsYou need a reasonably organized knowledge base. It doesn’t have to be perfect; even a well-maintained SharePoint or Google Drive is a solid starting point. Ongoing maintenanceAs your documents and processes change, the assistant’s knowledge base needs periodic updates. This can be automated in most cases. The Impact Teams that deploy internal AI assistants consistently report finding answers in seconds rather than minutes, reducing dependence on “knowledge gatekeepers,” faster onboarding for new hires, and more consistent application of company policies. The assistant doesn’t replace your team’s expertise; it makes it accessible to everyone, instantly. Want to explore what a custom AI assistant could do for your team? Let’s build it together.
The True Cost of Manual Data Entry in 2026
Previous Post Manual data entry feels cheap because it doesn’t show up as a line item on your budget. There’s no invoice for it. No subscription fee. It’s just… people doing work. But that’s exactly what makes it so expensive. Opportunity Cost: The Biggest Line Item The most expensive part of manual data entry isn’t the labor or the errors, it’s what your team isn’t doing while they’re busy typing. Your sales rep spending an hour logging CRM data is an hour they’re not selling. Your finance person spending the morning on invoice entry is a morning they’re not analyzing cash flow. Your operations manager updating spreadsheets is an operations manager, not optimizing your supply chain. This is the invisible cost: the growth that doesn’t happen because your best people are stuck on low-value work. The cost of fixing a single data entry error varies widely, but estimates typically range from €10 to €100+, depending on the industry and impact. Even at the low end, that’s an additional €1,000–€10,000 per year in error-related costs. The Hidden Costs Let’s do some simple math. Say you have one team member spending 2 hours per day on data entry tasks, copying information between systems, entering form responses, updating spreadsheets, logging emails into your CRM. Two hours per day is 10 hours per week. That’s roughly 500 hours per year. If that employee earns €25 per hour fully loaded (salary plus taxes, benefits, equipment), that’s €12,500 per year spent on typing data that a machine could handle. And that’s just one person. Most companies have this pattern repeated across multiple roles. Error Rates Add Up Humans make data entry errors at a rate of approximately 1%, and that’s for focused, experienced workers. For fatigued or distracted employees, it’s much higher. A 1% error rate across 10,000 entries means 100 errors. Each error requires detection (if it’s even caught), investigation, and correction. Some errors have downstream consequences, such as a wrong invoice amount, a misrouted order, or a compliance violation. The cost of fixing a single data entry error varies widely, but estimates typically range from €10 to €100+, depending on the industry and impact. Even at the low end, that’s an additional €1,000–€10,000 per year in error-related costs. What Automation Actually Costs The alternative isn’t expensive. A Power Automate flow that replaces a manual data entry task might cost €500–€2,000 to set up and €50–€100/month to maintain. A custom Python script running on Azure might cost a few hundred to build and pennies per month to run. Compare that to €12,500+ per year in manual labor, and the ROI is obvious within the first month. The Mindset Shift Many companies treat data entry as “just part of the job.” And for a long time, it was. But in 2026, with the tools available today, choosing to do data entry manually is choosing to waste money. The question isn’t whether you can afford automation. It’s whether you can afford not to. Ready to calculate your real automation ROI? Book a free call and we’ll map it out.
How AI Document Processing Works (and Why You Need It)
Previous Post Every business runs on documents. Invoices, contracts, applications, receipts, reports, forms, they flow in constantly, and someone has to read them, extract the important information, and enter it into the right system. That “someone” doesn’t have to be a person anymore. What Is AI Document Processing? AI document processing uses artificial intelligence to automatically read, understand, and extract information from documents regardless of format, layout, or language. Unlike traditional OCR (optical character recognition), which simply converts images to text, modern AI document processing actually understands what it’s reading. It knows that “Total Due: €4,500” is an invoice amount, that “John Smith” in the signature block is the signatory, and that “Net 30” means payment terms. How It Works The process typically follows four stages. IngestionDocuments arrive from any source: email attachments, scanned files, uploaded forms, shared folders. The system accepts PDFs, images, Word documents, and even handwritten notes. ClassificationThe AI identifies the document type. Is this an invoice? A contract? An application? A receipt? This determines how it’s processed next. ExtractionThe AI pulls out key data fields based on the document type. For an invoice, that means vendor name, invoice number, line items, total amount, and due date. For a contract, it might be the parties involved, effective dates, terms, and obligations. Validation and outputExtracted data is validated against your business rules (does this vendor exist in our system? Is the amount within the expected range?), then delivered to your database, ERP, CRM, or any other destination. The ROI Is Hard to Ignore Manual document processing costs between €5–€15 per document when you factor in labor time, error correction, and delays. AI processing brings that down to cents per document while running around the clock with consistent accuracy. For a company processing 500 documents per month, the savings can easily reach five figures annually, not counting the value of faster turnaround times and reduced errors. What Makes It Different from OCR? Traditional OCR reads characters. AI document processing reads meaning. OCR might extract the text “Net 30” from a document. AI document processing understands that this represents 30-day payment terms and maps it to the correct field in your system. The difference is the gap between raw text and actionable, structured data. Modern AI models can also handle messy, inconsistent documents, different layouts, varying formats, handwritten annotations, and even documents in multiple languages. Real-World Applications Accounts payableInvoices from hundreds of vendors, each with different formats, are automatically processed and entered into your accounting system. Insurance claimsClaim forms, supporting documents, and medical records are automatically parsed, classified, and routed for assessment. Legal document reviewContracts scanned for key clauses, obligations, deadlines, and risk factors in minutes instead of hours. HR and recruitmentResumes are parsed for skills, experience, and qualifications, then automatically scored against job requirements. LogisticsBills of lading, customs declarations, and shipping documents are processed automatically across international supply chains. How EntryHub Builds It We build AI document processing pipelines using a combination of Azure Cognitive Services, Azure OpenAI, and custom Python logic. The result is a system tailored to your specific document types, business rules, and destination systems. We handle everything from document ingestion to data delivery, including exception handling for documents that need human review. Processing documents manually in 2026? Let’s fix that.
Getting Started with Azure Functions for Business Automation
Previous PostNext Post If you’ve outgrown simple automation tools and need something more powerful, flexible, and scalable, Azure Functions is where things get interesting. What Are Azure Functions? Azure Functions is Microsoft’s serverless computing platform. In plain terms, it lets you run code in the cloud without worrying about servers, infrastructure, or scaling. You write a function, deploy it, and Azure handles everything else. Your function can be triggered by virtually anything: an HTTP request, a timer schedule, a new file in storage, a message in a queue, a database change, or an event from another Azure service. You write the code (Python, C#, JavaScript, or others), define the trigger, and Azure runs it automatically whenever the trigger fires. You only pay for the time your code actually runs. Why Businesses Love Azure Functions No infrastructure to manage. You don’t need to set up servers, install software, or handle updates. Azure manages the runtime environment entirely. Scales automatically. Whether your function runs once a day or a thousand times per minute, Azure scales it up and down dynamically. You never have to plan for capacity. Cost-efficient. The consumption plan gives you the first million executions per month for free. After that, you pay fractions of a cent per execution. For most business automation use cases, the cost is negligible. Integrates with everything. Azure Functions connect natively with Azure Storage, Cosmos DB, Service Bus, Event Grid, and dozens of other Azure services. They also work seamlessly with Power Automate, Logic Apps, and external APIs. Real Business Use Cases Scheduled data processing A Python function runs every night at 2 AM, pulls data from three APIs, transforms and consolidates it, and writes the result to a database. Your team wakes up to fresh, clean data every morning. Real-time file processing When a new file lands in Azure Blob Storage (uploaded by a partner, a scanner, or another system), a function triggers immediately, processes the file (extracts data, converts format, validates content), and routes the result to the appropriate system. Webhook receiver A third-party service (Stripe, HubSpot, Shopify) sends a webhook when something happens. Your function receives it, processes the payload, and updates your internal systems in real time. AI processing pipeline A function receives a document, sends it to Azure Cognitive Services or Azure OpenAI for analysis, processes the results, and stores the output — all serverless, all automatic. Backend for Power Automate Power Automate handles the trigger and routing, then calls an Azure Function via HTTP for the complex processing logic. Best of both worlds. Getting Started: The Basics To create your first Azure Function, you need an Azure account (free tier available), Azure Functions Core Tools installed locally for development, and a code editor like VS Code with the Azure Functions extension. The development cycle is straightforward: write your function locally, test it on your machine, deploy it to Azure with a single command, and configure the trigger. Microsoft’s documentation walks through every step. When to Use Azure Functions vs. Power Automate Use Power Automate when the automation connects Microsoft 365 apps and the logic is straightforward. Use Azure Functions when you need custom code, complex processing, or integration with non-Microsoft systems at scale. In practice, many of our best solutions combine both: Power Automate for the trigger and user-facing layer, Azure Functions for the processing engine. The EntryHub Approach We build Azure Functions in Python for clients who need automation that goes beyond what low-code tools can handle. Whether it’s processing thousands of documents, running AI models, or orchestrating complex multi-system workflows. We design, build, deploy, and monitor the entire solution. Ready to explore what Azure Functions can do for your business? Let’s talk.
How to Automate Invoice Processing with Power Automate
Previous PostNext Post Invoice processing is one of the most common and most tedious tasks in any business. Invoices arrive by email, someone manually enters the details, routes them for approval, tracks the status, and logs everything in a spreadsheet or accounting system. It’s slow, error-prone, and a poor use of your team’s time. Here’s how Power Automate can handle it instead. The Typical Manual Process Let’s look at what most companies do today: An invoice arrives as an email attachment. Someone downloads it, opens the PDF, and manually types the vendor name, invoice number, amount, due date, and line items into a spreadsheet or ERP system. Then they forward it to a manager for approval, often via email. The manager might approve it immediately, or it sits in their inbox for days. Once approved, someone updates the status, schedules the payment, and files the document. This process typically takes 15–30 minutes per invoice. Multiply that by dozens or hundreds of invoices per month, and you’re looking at a significant time drain. The Automated Workflow Here’s how this looks with Power Automate: Step 1: Trigger – Invoice arrives by email The flow monitors a shared mailbox or specific inbox folder. When an email with an attachment arrives (filtered by subject line, sender, or attachment type), the flow triggers automatically. Step 2: Extract data using AI Builder Power Automate’s AI Builder includes a pre-built invoice processing model. It reads the PDF attachment and extracts key fields: vendor name, invoice number, date, total amount, line items, and more. No manual data entry required. Step 3: Log the data The extracted information is automatically written to a SharePoint list, Excel file, or Dataverse table, whatever your team uses for tracking. Step 4: Route for approval Based on the invoice amount or department, the flow sends an approval request to the right manager via Teams, email, or the Approvals app. The approver can review the details and approve or reject with a single click. Step 5: Update status and notify Once approved, the flow updates the invoice status in your tracking system, sends a confirmation to the relevant parties, and optionally moves the file to an “Approved” folder in SharePoint. Step 6: Handle exceptions If AI Builder can’t extract certain fields with confidence, or if the invoice is rejected, the flow routes it to a human for review with all available data pre-filled. What You Need to Get Started To build this flow, you’ll need a Microsoft 365 subscription with Power Automate (included in most business plans), AI Builder credits (included in Power Automate Premium or available as an add-on), a shared mailbox or dedicated inbox for invoices, and a SharePoint list or Dataverse table for tracking. Real-World Results Companies that automate invoice processing typically see processing time drop by 70–80%, data entry errors virtually eliminated, approval cycle times cut from days to hours, and full audit trails with zero manual effort. Beyond the Basics The workflow above covers the fundamentals, but you can extend it further. You can connect the approved invoices directly to your accounting software via API. You can set up recurring reports that summarize monthly invoice volumes and amounts. You can build a Power BI dashboard that gives finance teams real-time visibility. The possibilities expand with each integration. DIY or Done For You? If you have someone on your team familiar with Power Automate, the basic version of this flow is achievable in-house. But if you want AI extraction, complex routing logic, error handling, and integration with your accounting system, that’s where a specialized partner makes the difference. Want us to build this for you? Get in touch, we can have it running within a week.
Power Automate vs. Python: Which Should You Use for Automation?
Previous PostNext Post When it comes to automating business processes, two tools come up again and again: Microsoft Power Automate and Python. Both are powerful. Both can save your team hours of manual work. But they’re fundamentally different tools designed for different situations. So which one should you use? Let’s break it down. Power Automate: The Low-Code Powerhouse Power Automate is Microsoft’s automation platform built into the Microsoft 365 ecosystem. It uses a visual, drag-and-drop interface to create automated workflows called “flows” that connect your apps and services. Where Power Automate shines: It’s the natural choice when your workflow lives inside the Microsoft ecosystem. Connecting Outlook to SharePoint, automating Teams notifications based on form submissions, and creating approval chains in Dynamics 365. These are Power Automate’s bread and butter. It’s also great for business users who need to build and maintain their own automations. The visual interface means no coding is required for standard flows, and the hundreds of built-in connectors make integration straightforward. Power Automate also includes desktop flows (RPA) that can automate legacy applications by simulating mouse clicks and keyboard inputs, useful when APIs aren’t available. Where it struggles: Complex data transformations, heavy computational logic, web scraping, working with files at scale, or anything that requires custom algorithms. Power Automate’s expression language is limited compared to a full programming language, and debugging complex flows can be frustrating. Power Automate is Microsoft’s automation platform built into the Microsoft 365 ecosystem. It uses a visual, drag-and-drop interface to create automated workflows called “flows” that connect your apps and services. It’s the natural choice when your workflow lives inside the Microsoft ecosystem. Connecting Outlook to SharePoint, automating Teams notifications based on form submissions, and creating approval chains in Dynamics 365. These are Power Automate’s bread and butter. It’s also great for business users who need to build and maintain their own automations. The visual interface means no coding is required for standard flows, and the hundreds of built-in connectors make integration straightforward. Power Automate also includes desktop flows (RPA) that can automate legacy applications by simulating mouse clicks and keyboard inputs, useful when APIs aren’t available. Complex data transformations, heavy computational logic, web scraping, working with files at scale, or anything that requires custom algorithms. Power Automate’s expression language is limited compared to a full programming language, and debugging complex flows can be frustrating. Python: The Swiss Army Knife Python is a general-purpose programming language that can do almost anything: data processing, web scraping, API integration, machine learning, file manipulation, and much more. Where Python shines: When you need full control. Python excels at complex data transformations, processing thousands of files, scraping websites, building custom APIs, running machine learning models, and handling edge cases with sophisticated error handling. It’s also the foundation of modern AI. If you’re building AI agents, working with language models, or processing unstructured data, Python is the tool of choice. Where it struggles: Python requires a developer to build and maintain. Non-technical team members can’t easily modify or troubleshoot Python scripts. It also needs a runtime environment, such as a server, cloud function, or container, which adds infrastructure complexity. Python is a general-purpose programming language that can do almost anything: data processing, web scraping, API integration, machine learning, file manipulation, and much more. When you need full control. Python excels at complex data transformations, processing thousands of files, scraping websites, building custom APIs, running machine learning models, and handling edge cases with sophisticated error handling. It’s also the foundation of modern AI. If you’re building AI agents, working with language models, or processing unstructured data, Python is the tool of choice. Python requires a developer to build and maintain. Non-technical team members can’t easily modify or troubleshoot Python scripts. It also needs a runtime environment, such as a server, cloud function, or container, which adds infrastructure complexity. The Decision Framework Here’s a practical guide for choosing: Choose Power Automate when your workflow connects Microsoft 365 apps. The logic is straightforward (triggers, conditions, actions), business users need to maintain or monitor the flow, you need RPA for legacy desktop applications, or you want something running within hours. Choose Python when you’re processing large volumes of data or files, the logic involves complex calculations or transformations, you need web scraping or custom API integrations, you’re building AI/ML-powered automation, or you need maximum flexibility and control. Use both when you need Power Automate for the trigger and routing layer (email arrives, form submitted, file uploaded), but Python for the heavy processing logic. This is actually one of the most powerful patterns. Power Automate handles the Microsoft integration, and an Azure Function running Python handles the complex work. Our Recommendation Don’t think of it as “either/or.” Think of it as “which tool fits this specific task?” At EntryHub, we regularly combine both in a single solution. Power Automate provides the connective tissue within Microsoft 365, while Python handles the custom logic and AI capabilities. The best automation strategy uses the right tool for each job. Not sure which approach fits your workflow? Book a free discovery call and we’ll map it out together.
What Is an AI Agent and How Can It Help Your Business?
You’ve probably heard the term “AI agent” thrown around a lot lately. But what does it actually mean, and more importantly, what can it do for your business?
Let’s cut through the hype and get practical.
5 Business Tasks You Should Automate Right Now
Every business has those repetitive, soul-crushing tasks that eat up hours every week but never actually move the needle. The kind of work that makes your best people feel like they’re stuck on a hamster wheel.
The good news? Most of these tasks can be fully automated. Not in some far-off future but right now, with tools that already exist.