Why am I reading this?
Large language models are here, reinventing the competitive landscape on which your company must now do battle. So, how are you going to win, and if you are not already leading the field, what time do you have left to figure this out?
Businesses able to weaponise their AI with the necessary human and technological resources are already closing in on victory, and if you don’t have big enough guns, you may already have lost, and, if so, may not even yet know this has happened, or how.
And, because AI models are only as good as the data they are trained on, even if you try to catch up, you may be wasting your time entirely if your AI training data isn’t clean enough to contend in the first place.
The path to competitively superior AI productivity runs directly through your modelling data's quality
The more readily machine-readable the work of your teams becomes, the faster your AI will conjure what your organisation can do to speed it to new efficiencies, learning, superior collective intelligence and fatter margins.
As not all action on AI is created equal, the following solution describes how to use longstanding commercially proven publishing and editorial methods to generate the data to deliver clean, reliable AI repeatably.
Kicking off by adhering to the tight prescriptions of Amazon’s Working Backwards method, and embedding such editing practices into AI-model training, organisations can reduce risk further in building dependable, trustworthy AI systems.
The following illustrates the simple steps professional editors would use as guidelines to ensure that the data put to work is of high quality, unbiased, and transparent to all stakeholders involved in AI modelling.
Work backwards the Amazon way
Amazon’s “working backwards” approach ensures its products are built to ensure they will deliver value to customers.
Instead of starting with a product idea, Amazon teams begin by writing a press release outlining a product’s key features, its value proposition, and how it will positively improve customers’ lives.
Coupled with a frequently asked questions (FAQ) document that answers all the likely questions this release must address, working backwards to put customers’ needs before building products to meet them, reduces the risk of creating goods that don’t meet customer expectations and allows Amazon to innovate more effectively.
What is data cleaning and why is it crucial in AI?
Following the simple rule of garbage in, garbage out (GIGO), data cleaning is crucial in AI because if data is dirty, is noisy or contains errors, an AI model’s predictions or decisions will be less accurate, and less reliable.
When data is collected from various sources, it may be in different formats, and if not cleaned to give it absolute consistency, dirty data can introduce bias, which can also lead to unfair or discriminatory predictions.
How much productivity lies hidden in what your business knows?
Every business knows more than it can express.
And, although this awareness exists in the qualitative, soft data which resides in people’s heads, when this can be captured effectively, organisational knowledge – the holy grail of AI – can be mapped by using knowledge graphs.
These must have an organising principle so that a user – or a computer system – can reason about the underlying data.
Organising principles, reasoning, and qualitative knowledge discovery might seem abstract and complicated at first.
And if the idea of building a workplace’s collective intelligence by capturing its knowledge, let alone that of a customer base, may sound chaotic and messy, it is.
Yet, making sense from a cacophony of competing voices and opinions is no different to the challenge faced by news organisations reporting around the world every day.
That is, in professional publishing there exists a highly evolved, long-established global content-management precedent, more than fit to address this challenge.
Once captured, knowledge graphs then become a rich index over data to provide the means of curation, much like a skilled librarian recommends books and journals to a researcher.
And getting this and what it leads to right in your business by using the following three straightforward disciplines can deliver the most powerful force for organising knowledge capital ever invented.
Discipline one: Create your editorial style guide
In professional publishing, an editorial style guide is an essential tool of efficiency, which can save time for writers and editors – and make life easier, less jarring and more predictable for readers – as its goal is to create the certainty that all published content is consistent, accurate, and easy to comprehend.
It ensures that written content adheres to a standardised set of guidelines, and a house style guide will typically include rules governing:
- Grammar and punctuation
- Spelling and capitalisation
- Names, labels, abbreviations and acronyms
- Numbers and dates
- Citations and references
- Style and tone
- Page or document structure
Training your people in how to adhere to a style guide can train your AI model over time with an efficiency and reduction of risk other methods might not achieve at all.
Discipline two: Identify and install your AI-training operation's high priest of content creation
In professional publishing, commissioning editors, often experts in their particular field, are responsible for ensuring that what gets published consistently meets the needs of its target audience.
Commissioning editors specify what gets written and who writes it. They work with and review a writer’s drafts, checking for accuracy and adherence to guidelines, until it is ready, at which point that content is then handed over to the production team, including layout artists and proofreaders.
Because it determines what content feeds your AI operation, this oversight function is likely therefore also critical in the consistent corporate delivery of useful, usable AI.
Discipline three: Establish your AI-training content-checking regime
In professional publishing, sub-editors – also known as copy editors – are responsible for the final checking and preparation of written content for publication.
They cut, rewrite and check for potential legal, libel or copyright violations to ensure that content adheres to necessary guidelines.
And, as the pickiest proofreaders within an editorial team, they are responsible for making sure that what gets written makes sense, is accurate and complies with the rules established in the style guide, and for ensuring that data production stays on schedule.
Build sense and fine-tuning to advance the capture of intelligence from collaborative workflows
Modern, iterative, collaborative workflows certainly introduce new challenges for AI data modeling.
With some workplace knowledge being explicit – written documents, databases – and some of the most important data being tacit, personal and previously unarticulated – expertise, experience, insight – the diversity represented by different teams or departments, each working with its own datasets and tools, can make it difficult to get a complete picture of what an organisation knows.
This inherent messiness must also be addressed by rigorous and consistent data cleaning within an organisation’s AI data-modelling strategy.
In summary, take these six steps to bed down your AI-modelling processes
- Pick your team of authors, to ensure the best minds across the business are engaged in framing and creating the collective intelligence on which the organisation’s AI is to be modelled, and from which it will learn.
- Work backwards with your team to articulate your first AI mission and how and why it will succeed.
- Select your “commissioning editor” – a highly literate senior manager with both broad and deep knowledge of your business – with trust across it, from the CEO, board and/or senior leadership team.
- Set up your style guide in consultation with your data scientists to ensure the structures of the key reports generated by your team meet critical continuing AI machine-learning needs.
- In line with your primary overarching business analysis – see beneath – decide, define and create the most important internal AI-assisted reports from which the AI model of your business is to learn.
- Train your team in how to create future documentation reliably in accordance with the style guide.
Conduct an overarching business analysis for mission-critical AI-model training
Here are some guiding suggestions to put to work in building your analytical framework, based on:
- PESTLE analysis
- Porter’s Five Forces
- Kaplan and Norton’s Balanced Scorecard
- Alexander Osterwalder’s Business Model Canvas
- Scenario planning
- A marketing SWOT: strengths, weaknesses, opportunities, threats
See appendix beneath for greater detail on each of these frameworks.
Conclusion
Making a workplace’s collective intelligence work more effectively to improve its ways of working in ways simply never previously possible before AI may take some new thinking and planning.
Yet, in preparing its documentation across the organisation in line with the needs of the style guide and under the control and careful checking of the commissioning editor, and those responsible for checking that what gets fed into the AI model is what was intended, the company that prepares its AI model with such discipline should find itself far better prepared for whatever the future and its possibly less well prepared competition throws at it.
As such, establishing this entirely new discipline to ensure the way it captures what is known by those working within it can become the defining hallmark of continuous improvement in your organisation.
And, by employing these straightforward practices, it can be your business that can trumpet how it will change your industry’s customers’ experience and expectations before others realise what is happening to them.
Act now
Now, let’s set up a workshop in which we can begin to engage and direct the intelligence of your team to get these essential disciplines working to achieve ever greater productivity for your business.
APPENDIX
Conduct an overarching business analysis
First, it will do your organisation no harm for its people to understand its operating models better, so you might wish to articulate these by going both broad and deep to articulate what your business knows, needs and cleans in its training data.
The aim of such analyses is to guide it in understanding its future. But here, the point is to provide the framework on which the AI can begin to interpret a new range of such possible futures.
While macro environmental forces influence the operations of all organisations in a general fashion, there are also more specific sets of forces within industries that have direct and powerful effects. To understand these, industry-specific analysis is needed.
As referred to above, then, here are some guiding suggestions to put to work in building your analytical framework, based on:
- PESTLE analysis
- Porter’s Five Forces
- Kaplan and Norton’s Balanced Scorecard
- Alexander Osterwalder’s Business Model Canvas
- Scenario planning
- A marketing SWOT: strengths, weaknesses, opportunities, threats
PESTLE analysis: Political, Economic, Socio-cultural, Technological, Legal and regulatory, the natural Environment
Strategic thinkers need to address a number of industry-specific questions. In their 2008 book, Evaluating a Company’s External Environment’, Crafting and Executing Strategy: The Quest for Competitive Advantage, authors Thompson, Strickland & Gamble identify seven:
- What are the industry’s dominant economic features?
- What is competition like, and how strong are each of the competitive forces?
- What is causing the industry’s competitive structure and business environment to change?
- Which companies are in the strongest and weakest positions?
- What strategic moves are rivals likely to make next?
- What are the key factors for competitive success?
- Is the industry attractive, and what are the prospects for above average profitability?
Michael Porter’s Five Forces of Competition
Porter’s Five Forces of Competitive Position Analysis was developed in 1979 by Michael E Porter of Harvard Business School as a simple framework for assessing and evaluating the competitive strength and position of a business organisation.
It gives insights into the competitive landscape, market dynamics, and potential areas of vulnerability or opportunity.
While it’s typically used for business strategy, you can adapt it to gain valuable insights about your organisation.
The Five Forces are:
Supplier Power: Examination of this force will reveal how much control your suppliers have over your organisation.
By assessing your relationships with suppliers, you can understand factors such as the availability of key resources, the stability of your supply chain, and potential cost pressures. You can use this to help you plan for supply chain risks and to negotiate better terms with suppliers.
Buyer Power: This force focuses on how much influence your customers have on your business.
Understanding buyer power helps you evaluate customer satisfaction, loyalty, and the ability to set prices. It can guide you in improving customer relations, tailoring products/services to their needs, and setting competitive prices.
Competitive Rivalry: This force looks at the intensity of competition within your industry. By analysing your competitors and their strategies, you can identify areas in which your organisation needs to improve, differentiate itself, or seize new opportunities.
This will also help you to evaluate market saturation and potential scope for growth.
Threat of New Entrants: This force explores how easy or difficult it is for new competitors to enter your industry or market. If barriers to entry are low, you might need to fortify your position through innovation, branding, or customer loyalty. If barriers are high, you can maintain a stronger competitive advantage.
Threat of Substitutes: Substitutes are alternative products or services that could replace yours, but are not always obvious, as organisations in one industry may be competing with those in other industries that produce substitute products that satisfy similar consumer needs, competing for the same dollar, but which may differ in specific characteristic. Understanding this force can help you anticipate changes in customer preferences and market dynamics, enabling you to work on product differentiation or to find new ways to add value to minimise the risk of substitution.
Kaplan and Norton’s Balanced Scorecard
When training your AI model, it’s essential to provide a diverse dataset that represents a holistic view of your organisation’s performance. The Balanced Scorecard is such a strategic performance management framework that helps organisations monitor and measure their performance from various perspectives, including financial, customer, internal processes, and learning and growth, including both historical and real-time data, where available.
The model should be trained to recognise patterns, anomalies, and correlations within these documents to provide a comprehensive assessment of your organisation’s health based on the Balanced Scorecard framework.
To train an AI model to understand and evaluate the health of your organisation by using the Balanced Scorecard, you might consider using the following types of documents and reports:
- Reports providing data on the financial health of the organisation, including profit and loss statements, balance sheets, cash flow statements, and budget reports.
- Customer satisfaction surveys, feedback and customer service reports can be valuable for understanding customer perspectives, sentiment and trends to assess customer satisfaction and loyalty.
- Internal process pocuments, such as process maps, efficiency reports, and operational performance metrics, to help AI models evaluate the efficiency and effectiveness of internal operations, to assess how well the organisation is delivering its products or services.
- Employee training and development records addressing learning and growth metrics, such as employee training records, development plans, and performance appraisals, to provide insights into the organisation’s efforts to enhance its workforce and its impact on overall organisational growth.
- Strategy and objectives documents, including those outlining the organisation’s strategic goals, objectives, and key performance indicators (KPIs) are crucial for training the AI model to align with the organisation’s strategic direction and in evaluating performance against strategic targets.
- Competitor analysis reports can also be valuable for benchmarking your organisation’s performance against industry peers.
- Market and industry trends related to market and industry trends, such as market research reports, can help the AI model understand the external factors affecting your organisation.
- Regulatory compliance documents and those related to adherence to industry standards are critical, especially in regulated industries, to evaluate conformance with relevant laws and regulations.
The Business Model Canvas
Alexander Osterwalder’s Business Model Canvas is a strategic management tool that helps organisations visualise and analyse their business models to identify key components and activities within a business model that contribute to an organisation’s overall success.
The canvas consists of nine building blocks, each of which can contribute to various organisational outputs.
Here’s a list of these building blocks and some of the outputs they might influence:
- Customer segments: Who are your customers?
- Value propositions: What value do you offer to your customers?
- Channels: How do you reach your customers?
- Customer relationships: What kind of relationships do you have with your customers? (Are sales one-off?/ repeat?/ serial?)
- Revenue streams: How do you generate income?
- Key resources: What key resources do you need to operate your business?
- Key activities: What key activities do you need to perform to deliver your value proposition?
- Key partnerships: Who are your key partners?
- Cost structure: What must you spend money on?
The organisational outputs can vary significantly depending on the specific business model and its goals. Outputs could include financial metrics (revenue, profit), customer-related metrics (satisfaction, retention), operational efficiency measures, and more.
In addition to these direct outputs, key activities can also lead to indirect outputs such as:
- Increased innovation capacity
- Improved operational efficiency
- Enhanced market position
- Greater social and environmental impact
By understanding the relationship between activities and outputs, organisations can make more informed decisions about how to allocate resources and achieve their goals.
SWOT analysis
A corporate marketing SWOT analysis is a strategic planning tool that helps businesses identify their strengths, weaknesses, opportunities, and threats. It is a valuable tool for understanding the current state of a business and in developing strategies to achieve its marketing goals.
Strengths are the internal factors that give a business an advantage over its competitors. Examples of strengths include a strong brand reputation, a loyal customer base, or a unique product or service offering.
Weaknesses are the internal factors that hinder a business’s performance. Examples of weaknesses include a limited budget, a lack of brand awareness, or outdated technology.
Opportunities are the external factors that could benefit a business. Examples of opportunities include new market trends, changes in consumer behavior, or the failure to launch a new comparable product or service by a competitor.
Threats are the external factors that could harm a business. Examples of threats include new competition, economic downturns, or changes in government regulations.
By understanding its strengths, weaknesses, opportunities, and threats, a business can develop marketing strategies that leverage its strengths, improve its weaknesses, capitalise on opportunities, and mitigate threats.
Scenario planning
Scenario planning is a strategic tool used in corporate strategy development. It involves envisioning and evaluating multiple plausible future scenarios to inform decision-making. It embraces uncertainty and complexity, enabling organisations to anticipate and adapt to various potential outcomes. It fosters flexibility, resilience, and creative thinking, allowing businesses to navigate dynamic environments and seize opportunities while mitigating risks.
In practice, strategic managers typically have a likely scenario in their head while they are developing strategy.
However, also in practice, the future’s imagined scenarios never work out exactly as may have been imagined – they will usually be worse or better to some degree.
The point of developing alternative scenarios is to trigger discussion and thought in the strategic management team about:
- Which critical issues need to be monitored so that if events in the future do turn out to be significantly different to those imagined, and how can the organisation respond?
- How to develop capabilities, commit resources or even develop contingency plans for, “What will we do if…?”
Closing up
By planning with rigour and checking that the AI-training data generated is exactly what was intended, the company that prepares its AI model with such discipline should find itself far more competitively comfortable than those against which it competes whose businesses have given the quality of the data on which they compete no such thought.
Act now
Contact me to set up a workshop in which we can begin to engage and direct the intelligence of your team to get the essential disciplines described above working to achieve ever greater AI-driven productivity for your business.
Graham Lauren
0416 171724
cloudcitizenx@icloud.com
Thank you for reading.