Intelligent Document Processing | Build vs Buy
The expectation that AI will disrupt industry triggered a hiring drive. Makes sense. You need people to understand and work with evolving technology.
But, the predicted disruption isn’t happening as fast as forecast. It means companies are being forced to let people go or find projects to justify salaries.
Project idea… let’s build our own intelligent document processing platform and save our company some money. How hard can it be?
Depending on your decision, our RFP template could prove handy.
“Why should we buy your product, when we have a team of developers who could build one for us?”
It’s a fair question.
Short answer. It’s more than a spreadsheet, some docs, and a data capture tool. An IDP solution manages the entire document processing lifecycle, end to end. From document understanding to workflows and communication. Extensions and integrations. Reporting and insights.
Build vs buy questions to ask yourself
Look at your options, the pros and cons, and then decide which will deliver the best ROI, performance, and fastest time to value.
Ask yourself the following questions…
- Is building an IDP solution a strategic priority for our business?
- Can we cost-effectively build our own solution?
- Do we have the internal resources available?
- Do we have the technical expertise in-house needed to build?
- Can we build and deploy within three months?
- Are we able to maintain the long-term solution?
- Can our solution adapt to future needs and technological advancements?
Yes. It’s a no brainer. Build your own IDP solution.
No, but you’re tempted to give it a go, here are the features you’ll need to develop…
IDP platform features for end-to-end document automation
Functionality you’ll need to build, maintain, and improve…
- Document ingestion and flow – you’ll need to build an infrastructure that can ingest documents from multiple channels – email, scanners, EDI/e-invoicing, SFTP, APIs, etc., with filtering and pre-processing.
- Verification – build a UI for data capture verification and correction, with features to manage complex tasks such as extracting complex line items from 30+ page documents. And access control for different users.
- Continuous learning – build a robust process for data collection, training, and comparing data, to ensure ongoing accuracy.
- Structured formats – develop separate processes to handle computer-readable file formats – JSON, XML, UBL, etc. Alongside traditional documents.
- Business logic – the infrastructure you build must be able to apply business rules and logic for every document processed to ensure only valid transactions end up in your downstream system.
- Data matching – you’ll need to build functionality to automate cross-validation with your master data, e.g., vendor lists, entity lists.
- Communication with senders – implement automated communication features to handle the supplier back and forth. Automated document status notifications and an AI copilot to generate custom emails.
- Integration – establish seamless bi-directional and real-time integration with your downstream systems. For example, pulling the master data and posting the transactions to the downstream system of record such as SAP, Coupa, NetSuite, etc.
- Archiving – you’ll need to pay for and manage a searchable archive for processed documents.
- Security – each step of the transactional process demands a high level of security, because of the nature of the data.
Building your own IDP solution
Now is an ideal time to consider building your own IDP solution. The market is flooded with publicly available large language models and supporting APIs.
Building your own intelligent document processing means each component will have to be built from the ground up or outsourced to third party providers. It’ll be expensive, will need a large team, and time. Lots of time. To build, maintain, and improve.
If you’re a large business with hundreds of developers able to shift their focus from current tasks and commit time to building your IDP solution, excellent. But I’d hazard a guess, and say that’s rare.
Pros of building your own IDP solution
- You build it, you’re in control – features. Integrations. Whatever you need to meet your strategic business goals. How many data servers? Your choice. Which integrations? SAP, Coupa, NetSuite… Take your pick.
- Customize to your unique business processes – I’m not talking just features. The whole thing developed to align with your workflows and data structure. A solution that evolves as your business grows.
Cons of building your own IDP solution
- It’ll cost you $$$ – hiring AI experts, buying advanced technologies, development costs. It’ll be a cross-team project, so internal resources will take a hit. Plus, the cost of onboarding your team. Consider the impact on your product development if you divert your IT team to your IDP solution.
- It’ll eat up your time – developing, coding, planning, testing, integrating, refining.
- It’ll have to be compliant – you’ll be in control of the security and compliance functions. Do you have an in-house expert?
- It’ll need training data – you’ll need access to large volumes of diverse documents for training. Invoices, purchase orders, receipts, etc. The Rossum AI document processing platform has three training levels…
- Pretraining on large text collections to gain language understanding.
- Training on millions of transactional documents with rich annotations for universal document understanding.
- Continuous fine-tuning on the customer’s task.
- It’ll need ongoing maintenance – you can’t build and walk away. Ongoing updates, bug fixes, new integrations, new processes to tackle, and more.
- It’ll have no support network – You’ll be the only company using your custom-built IDP solution. No forums to ask for a bug fix. No sharing best practices. No online tutorials.
If you buy an IDP solution
When you buy an IDP solution, you can’t plug it in and seconds later see results. Any piece of advanced tech needs to be configured to your needs. Integrations. Team members onboarded. But, this will all be managed by your IDP vendor.
Pros of buying an IDP solution
- Implementation will be fast – plug-and-play solutions can be deployed quickly and easily, with minimal disruption to your workflow. The platform will be customized to your specific requirements, trained on a huge dataset, and onboarding will be handled by the vendor.
- You’re buying experts too – benefit from the vendor’s expertise in AI and IDP technology. They’ve built the platform. Tested it. Improved it. Added features. Maintained it. They understand your use case. They will support you.
- It’ll scale with your business – your business will grow. Your IDP solution has to grow too. An increase in document volume, new formats, use cases, regulations. A pre-built IDP solution will be scalable.
- You’ll have maintenance cover and updates – working with an IDP vendor, you’ll receive maintenance and regular updates. There’ll be ongoing improvements, to ensure the solution meets your evolving needs.
Cons of buying an IDP solution
- Customization blockers – the level of customization in a pre-built IDP solution varies according to the platform. If your use case is low volume, single format documents, an out-of-the-box solution will tick all the boxes. If you need flexibility in document formats and volume, and the ability to set up new processes, limited customization could be an issue.
- Integration issues – introducing any new technology to your tech stack can be tricky. For instance, an out-of-the-box solution may prove more problematic than a custom-built solution. Best practice, choose an IDP platform that seamlessly integrates with your existing tech stack.
- A vendor that won’t listen – you’re looking for a partner. An IDP vendor that’ll work with you. Understand your use cases. Anticipate your specific needs and evolve the solution with you.
Conclusion | Build vs buy
So. Build or buy? The decision you arrive at will depend on your business needs. Use cases and processes. IT team. Budget.
For most businesses, buying a pre-built IDP solution is the faster, more cost-effective, and lower-risk option. You’ll work with a robust, scalable solution with ongoing support from experts.
Remember my project idea – build an IDP solution if it could save money.
I’ve changed my mind.
Guessing you might need this. How to write an RFP for intelligent document processing. At Rossum, we’ve several years of expertise in developing, implementing, and refining business processes. We’ve seen our fair share of RFPs and we’d like to share what we’ve learned. This post takes you through the processes and gives you an RFP template with suggested questions to ask your chosen IDP vendor.